WO2006093452A1

WO2006093452A1 - Method for the life prediction of vehicle components

Info

Publication number: WO2006093452A1
Application number: PCT/SE2006/000172
Authority: WO
Inventors: Jörgen ANDERSSON; Fredrik Öijer; Inge Johansson
Original assignee: Volvo Lastvagnar Ab
Priority date: 2005-03-02
Filing date: 2006-02-08
Publication date: 2006-09-08
Also published as: SE0500496L

Abstract

A method for the life prediction of a plurality of vehicle components comprising the following method steps; estimation of the surface roughness of the road upon which the vehicle is travelling, measurement of the speed of the vehicle and estimation of an individual wear index for each component from the parameters surface roughness and speed of the vehicle. The object of the invention is to be able to judge when a component on a vehicle is in need of service.

Description

TITLE

Method for the life prediction of vehicle components

TECHNICAL FIELD

The present invention relates to a method for the life prediction of vehicle components. In particular, it relates to the life prediction of vehicle components that have a life that is limited by repeated mechanical loading that can give rise to fatigue failure.

BACKGROUND ART

A vehicle is a complex system assembled of individual technical components. The fact that the components interact means that a fault in one component can result in other parts in the system being damaged. When this is the case, such resultant damage can be very expensive to repair. For this reason, a limited life is specified for certain components. In order to reduce the risk of consequential damage, the component is to be replaced when the limited life has expired, whether this is necessary or not. The calculated life is often dependent upon the expected cost of the consequential fault, in such a way that the greater the expected cost, the more conservative the calculation of the life. The life of a product can normally be described by a probability distribution that describes the probability of a fault in the component as a function of the utilization time for the product. By a conservative calculation of the life is meant that the life is determined in such a way that the probability of a fault arising in the component before the calculated life has been reached is very small. When the life is calculated in a conservative way, the actual life of a product is statistically greater than the calculated life. This results in increased costs for the owner of the vehicle, as products that are still in working order must be replaced.

In order to reduce the difference between the calculated life and actual life, different forms of life prediction for components can be utilized. By life prediction is meant that a measurement of a remaining period of time, distance or other similar quantity is determined from input data from a sensor mounted on the vehicle. In its simplest form, the life prediction can be carried out by utilizing information about the distance that the vehicle has travelled since the most recent replacement of the component. An example of a component that utilizes such a life prediction is engine oil. The life of engine oil is often determined as a maximum number of kilometres. The engine oil is replaced at certain specified service intervals. In more advanced systems for predicting life, other types of sensor are utilized. For example, in US 6230554 a combination of sensors is utilized for estimating the status of a component from information about the temperature of the component, the gradient of the road, and the vehicle's load. From this input data, the status of the component and rate of change of the status of the component can be estimated. The status of the component can then be utilized to determine the need for a service or replacement of the component.

Even though previously-known systems for life prediction have improved the ability to determine the remaining life of components and thereby reduced the difference between the predicted life and actual life for certain types of components, there is still a need for systems and methods for the life prediction of vehicle components when the life of the components is limited by repeated mechanical loading that can give rise to fatigue failure.

DISCLOSURE OF INVENTION

An object of the invention is to provide a method for the life prediction of vehicle components when the life of the components is limited by repeated mechanical loading that can give rise to fatigue failure. This object is achieved by the method defined in claim 1. According to the invention, the life of vehicle components is predicted by estimating the surface roughness of the road upon which the vehicle is travelling, measuring the speed of the vehicle and by estimating a wear index from the parameters surface roughness and speed of the vehicle. By wear index is meant a measurement of remaining life. This measurement can, for example, be expressed as a number between 0 and 1 , where 1 means that the component is completely new and 0 means that the component has no remaining life. It is advantageous if individual wear indices can be estimated for different components or for different groups of components.

The mechanical force on a component is dependent upon the surface roughness of the road upon which the vehicle is travelling and knowledge of the speed of the vehicle. This means that the magnitude of vibrations that arise as a result of the vehicle being driven over the surface is dependent upon the speed of the vehicle and upon the roughness of the road surface. As the mechanical force is dependent upon the said parameters, the life of the component is also dependent upon the same parameters. The life of the component can thus be determined empirically from input data about the speed and the roughness of the road surface.

According to a preferred embodiment of the invention, the roughness of the road surface is determined by measuring the vertical acceleration of a component attached to the vehicle. The vertical acceleration is suitably measured as close to the road as possible. For this purpose, a positioning of an accelerometer on the vehicle's wheel axle is particularly expedient. If the accelerometer is attached to the vehicle's wheel axle, vibrations that arise will not be filtered by the suspension that supports the wheel axle and by any other components on the vehicle, which makes it easier to obtain a correct estimation of the surface profile of the road.

According to yet another preferred embodiment of the invention, the surface roughness is made up of a random interference term and discrete transient irregularities. The random interference term is utilized to estimate small irregularities, that can suitably be described by a distribution function, see Figure 1. The distribution function describes the state φ of the roadway measured in [m³/rad] as a function of the angular frequency Ω of the roadway. The random interference term for the road surface roughness can often be described by the approximation Θ(Ω) = ΘoΩ^"w, where Θ_o is the state of the roadway when Ω = 1 rad/m, and w = the gradient of the distribution function when Ω = 1 rad/m. By a study of the distribution function of the random interference term as described above, the random interference term for the surface roughness can be described, for example by identification of the parameters Θo and w for the road. For further information about the representation of road surface roughness according to the model described above, refer to "The description of road surface roughness", CJ Dodds & J. D. Robson, Journal of Sound and Vibration (1973) no 31(2) pages 175 - 183. When the random interference term for the surface roughness has been determined, the information is utilized that describes the random interference term for the surface roughness, for example the parameters Θo and w, in order to determine how much damage is caused by travelling over a road of the type that has been identified. In commonly-used models, the roughness of the road surface is described for wavelengths between 60 m and 3 dm. For wavelengths of less than 3 dm, the vehicle's tyres constitute an adequate filter that prevents the road surface from affecting the vehicle's other components. The amount of damage to components in relation to the random interference term is stored in a database that is determined empirically or by means of simulations. The forces from the random interference term on a component mounted on the vehicle can be determined in a simple way by calculating the load on the different parts of the vehicle depending upon the magnitude and frequency distribution of the random interference term using a finite element model of the vehicle. With knowledge of the vehicle's patterns of movement, which are described by the random interference term, in addition to the part that is dependent upon transient phenomena, the stresses σ on the different components in the vehicle can be calculated. From knowledge of the stresses σ, part damage can be determined by modelling of stress and part damage. The tables are based on knowledge of the material characteristics of the components and usual diagrams that show the stress limit as a function of the number of repetitions.

The discrete transient irregularities can not be described as a statistical distribution, which is the case for small interferences that can be represented via a random interference term. In addition, the discrete transient interferences constitute a large load on the components due to the fact that they give rise to relatively large forces acting on the vehicle and its components.

For fatigue failure, it is theoretically the case that for stresses that are less than the fatigue limit, there can be an infinite number of load repetitions. For stresses over the fatigue limit, it is generally the case that the number of permitted load repetitions decreases logarithmically as a function of the amplitude of the stress.

In a preferred embodiment of the invention, it is not, however, assumed that there can be an infinite number of repetitions for stresses that are below the fatigue limit. Instead, the stress limit for stresses that are below the fatigue limit is approximated by the continuation of the straight line in a diagram of the stress limit as a function of the logarithm of the number of repetitions for stresses exceeding the fatigue limit.

If stresses above the fatigue limit can be identified, or at least if stresses above a certain size can be identified, an approximation of the estimated life of a component can be improved considerably. For this purpose, transient irregularities must be able to be identified from the measurement data from a sensor, preferably an accelerometer that is arranged to estimate the surface roughness of the road.

In order to be able to identify transient irregularities from measurement data from the sensor, it has been found that the signal must be filtered. In an embodiment of the invention, the filtering is carried out by means of so-called wavelet transforms, where a signal is described by means of a set of base functions, each of which has a magnitude that is determined from a set of input data. For a fuller description of how wavelet transforms are used, refer to Donoho, D.: Non-linear Wavelet Methods for Recovery of Signals, Densities, and Spectra from Indirect and Noisy Data, Proc. of Symposia in Applied Mathematics, Volume 47, 1993; Donoho, D. and Johnstone, K: Ideal Spatial Adaptation via Wavelet Shrinkage, Biometrika, 81:425-455, Dec. 1994; or Mallat, S.: A Wavelet Tour of Signal Processing, 2nd Edition, Academic Press, 1999.

In an embodiment of the invention, the category of base functions Daubechies4 is selected. The number of base functions is selected automatically dependent upon the number of measurement points. The different base functions each have different frequencies. By means of the filtering with the wavelet transform, the noise that is superimposed on the response from the transient irregularity can be eliminated. Once the base functions and their magnitude have been determined, the inverse transform of the transform result is carried out.

After the inverse transform of the acceleration signal has been carried out, the acceleration is described as a discrete set of wave packets. This set of wave packets can be compared with the response from the idealized transient irregularities, the form and size of which are known, after which an identified transient irregularity can be approximated by one or a combination of several of the idealized transient irregularities. The idealized transient irregularities can thus be described as transient states.

Figure 2 shows an example of a transient state resulting from a transient irregularity consisting of an idealized hole with a depth of 6 cm and a length of 80 cm. The forces from the idealized irregularities have been determined via a finite element model of the vehicle. Part damage can be determined by means of knowledge of the stress σ on each of the components arising as a result of an idealized irregularity. The part damage can be determined empirically or theoretically by means of a diagram of the fatigue limit as a function of the logarithm of the number of load repetitions. By approximating the transient irregularities with the idealized irregularities, a good estimate of the wear from transient irregularities and thereby a good estimate of the life of the components can be obtained.

In order to obtain an estimate of the life of the components, an accumulated wear index is continually updated by the effect of the random interference term on the wear index being integrated over the time that the vehicle is used and by the effect of the transient irregularities on the wear index being summed over the time that the vehicle is used. In addition, the accumulated wear index for each component can be compared with a limit value for the wear index, after which an error message is generated if the wear index exceeds a limit value that has been determined for the component.

BRIEF DESCRIPTION OF DRAWINGS

An embodiment of the invention will be described below in greater detail with reference to the attached drawings.

Figure 1 shows an example of a distribution function that describes a random interference term for the surface roughness of a road.

Figure 2 shows an example of a composite transient resulting from a transient irregularity.

Figure 3 shows schematically a system for the identification of the surface roughness of the road upon which a vehicle is travelling and for the determination of damage that has occurred in a set of components comprised in the vehicle. Figure 4 shows a block diagram for a system for the determination of damage that has occurred in a set of components comprised in the vehicle, together with an indication of the method steps that are carried out.

MODE(S) FOR CARRYING OUT THE INVENTION

Figure 3 shows schematically a system for the identification 1 of the surface roughness of the road upon which a vehicle is travelling and for the determination 2 of damage that has occurred in a set of components comprised in the vehicle. Together, the system for the identification of the surface roughness 1 and the system 2 for the determination of damage that has occurred constitute a system for the life prediction of vehicle components. By means of input data from a first block 3, vehicle data is identified that constitutes information about the components that are comprised in the vehicle that are of significance for the determination of damage that has occurred in components and thereby the life of components comprised in the vehicle. For example, the information can comprise information about what type of suspension has been selected and thereby information about how the suspension is affected at different loads. In addition, input data about axle load from the vehicle is generated from a second block 4. The axle load can be obtained from a sensor. It is also possible to assume that the axle load is the same, irrespective of the driving situation. This approximation is sufficiently accurate if the vehicle is not carrying a load that has a weight that is dependent upon what type of transportation is being carried out, for example if it is a crane lorry that carries the same load all the time, and also if the sensor that is to be utilized to generate input data for estimating the roughness of the road surface is located on a wheel axle that carries essentially the same load irrespective of the driving situation, which can be the case when a sensor is arranged on a front axle of a goods vehicle. Input data from the first and second blocks 3, 4 is used to generate the vehicle model that describes how forces are transmitted between the components in the vehicle when vibrations from the surface upon which the vehicle is being driven are transmitted from the road to the components in the vehicle. Modelling can be carried out by the vehicle being approximated by a finite element model of the vehicle. The model of the vehicle is stored in a third block 5. The model consists suitably of a finite element model of the vehicle. The stresses on components as a function of the appearance of the roadway can be determined via a finite element model. By the appearance of the roadway is meant the random interference term and also the transient irregularities.

The models in the third block 5 are utilized to determine transfer functions between a random interference term in the input signal from the sensor 6 that provides input data for estimating the surface roughness of the road and a part of the surface roughness of the road that can be described by a random interference term, which is carried out in a fourth block 7, and they are also utilized to determine transfer functions between transients in the input signal from the sensor 6 that provides input data for estimating the surface roughness of the road and transient irregularities in the road, which is carried out in a fifth block 8. The transfer function between measured acceleration and the appearance of the roadway for the random interference term can be determined empirically by input data from a test track with a known random interference term being used as an input signal, together with a signal from the sensor 6 measured during a drive on the same test track. The sensor 6 consists preferably of an accelerometer. The accelerometer is suitably attached to a component that is as close to the road as possible, that is preferably the wheel axle or any component that is permanently attached to the wheel axle. The transfer functions are determined for a set of different speeds and axle loads, after which a database with transfer functions can be stored or modelled as an algorithm in a computer onboard the vehicle. It has been found expedient to integrate the acceleration two times over the time of a measurement of the displacement, in order to be able to determine the transfer function with good numerical stability. The transfer function can be expressed as Tχy(f) = Pχy(f)/Pχχ(f), where Pχχ(f) is the power spectrum of the displacement and Pχy(f) is the cross spectrum of displacement and road profile. In order to calculate the transfer function for the random interference term, both input data and output data for the system must be known. Input data from a known road profile is therefore utilized as input data for a finite element model in order to calculate the acceleration at the position of the sensor. Thereafter, the transfer function can be determined.

The transfer function between transient irregularities and measured input signal from the sensor 6 is determined by the response from the sensor 6 for different idealized irregularities being determined via a finite element model with an irregularity with a specific appearance, a particular speed and a particular axle load as input data. Also in this case, a set of transfer functions for different speeds and axle loads is determined. The response from a sensor can be calculated by means of a finite element model, with a known road profile with an idealized transient surface roughness as input data.

For a specific vehicle, the transfer function for the part of the surface roughness that is described by a random interference term for an input signal, and the transfer function for transient parts of the input signal, can be directly represented in a memory module in a computer or microprocessor. The transfer functions can be represented in any way known to the skilled person, such as algorithms, models or a database. In this case, a model of the vehicle does not, of course, need to be stored in the computer. The first and third blocks are thus only needed for the derivation of valid transfer functions, which can be dependent upon the axle load. Input data from a load sensor can be used when the axle load can vary and the transfer functions are dependent upon the load.

The road profile can be estimated utilizing the transfer functions in the fourth and fifth blocks. This is carried out in a sixth block 9. In an embodiment of the invention, input data about measured vertical acceleration, which is obtained from an accelerometer which is preferably attached to a wheel axle, preferably the front axle, is utilized as input data from the sensor 6. In addition, input data about the vehicle's speed is obtained from a seventh block 10. The input signal from the accelerometer 6 is made up of a part that is described as a random interference term and a filtered part that corresponds to the transient irregularities. A road profile is identified from the part of the input signal that can be described with a random interference term. This part of the road profile can preferably be described by a distribution function in accordance with Θ(Ω) = Θ_OΩ^'W, where Θo is the state of the roadway when Ω = 1 rad/m and w = the gradient of the distribution function when Ω = 1 rad/m. The transient parts of the signal are preferably filtered by means of the utilization of wavelet functions in accordance with what was described above. Transient irregularities can thereafter be identified by comparison of the filtered signal and stored signal responses from idealized irregularities. The comparison of identified transients in the filtered signal and stored signal responses from idealized irregularities can be carried out in any way that is well known to the skilled person, for example by algorithms, models, or a database. After the comparison, the identified transients in the filtered signal are approximated by an idealized irregularity or a superposition of idealized irregularities. The surface roughness of the road is thus determined from the input signal from the accelerometer 6, information about the vehicle's speed from the eighth block 10 and, where applicable, information from a load sensor 4, which surface roughness of the road is preferably described as a random interference term and transient irregularities.

Once the surface roughness of the road has been determined in the sixth block 9, a calculation is carried out of a wear index in a system for the determination of damage that has occurred in vehicle components 2. For each component, there is a model of a relationship between the surface roughness of the road, the current speed of the vehicle and the effect on the component in the form of calculated wear. The surface roughness is described in accordance with what is stated above, by means of a random interference term and discrete transient irregularities. In order to obtain an estimate of the life of the components, an accumulated wear index is continually updated by the effect of the random interference term on the wear index being integrated over the time that the vehicle is in use and by the effect of the transient irregularities on the wear index being summed in a ninth block 11 over the time that the vehicle is in use. In addition, the accumulated wear index for each component can be compared with a limit value for the wear index, after which an error message is generated if the wear index exceeds a limit value that has been set for the component.

Figure 4 shows a block diagram for a system for the determination of damage that has occurred in a set of components comprised in the vehicle, together with an indication of the method steps that are carried out. In a first method step S10, the vertical acceleration z is recorded. In a second method step S20, the part of the road profile that can be described as a random interference term Φ(Ω) is identified by the utilization of the input signal z and the transfer function between the vertical acceleration and the random interference term. In a third method step S21, the input signal corresponding to the vertical acceleration is filtered through a wavelet filter, whereby random interference terms comprised in the signal are essentially eliminated. After the filtering, the output signal corresponds to the excitation levels of a number of base functions. A comparison with a number of reference irregularities ID is carried out in a fourth method step S22, whereby a measured transient irregularity is approximated with one or more known reference irregularities. In a fifth method step S30, part damage ΔS(Φ) is determined for the comprised components A₁B₁C₁D, etc, that are monitored. In a sixth method step S31 , part damage ΔS(ID) is determined for the comprised components A,B,C,D, etc, that are monitored. In the fifth and sixth method steps, the speed (v) and the load (L) are utilized as input data for the system. In a seventh method step S40, the part of the total damage S that originates from the random interference term for each of the components is continually updated. This is carried out, in principle, by means of a time integral of the part damage ΔS(Φ) or by summation of ΔS(Φ) multiplied by the interval of time for which the estimate of ΔS(Φ) applies. In an eighth method step S41 , the part of the total damage S that originates from transient irregularities for each of the components ΔS(ID) is continually updated. This is carried out by part damages from different transient irregularities being summed. In a ninth method step, the total damage S from the irregularities calculated in the seventh and eighth method steps is summed.

An embodiment of a method for determining damage that has occurred in a set of components comprised in a vehicle can thus, in general, be expressed as follows:

- recording of the vertical acceleration z (S10); - identification of the part of the road profile that is described as a random interference term Φ(Ω), by the utilization of the input signal z and the transfer function between the vertical acceleration and the random interference term (S20);

- filtering of the input signal corresponding to the vertical acceleration through a wavelet filter, whereby random interference terms comprised in the signal are essentially eliminated (S21);

- comparison with a number of reference irregularities ID is carried out, whereby a measured transient irregularity is approximated with one or more known reference irregularities (S22); - determination of part damage ΔS(Φ) from the random interference term for the comprised components (A₁B, C₁D) that are monitored (S30);

- determination of part damage ΔS(ID) from the transient irregularity for the comprised components (A₁B₁C₁D) that are monitored (S31); - continual updating of the part (S_Randm = JAS(Φ)dt) of the total damage S that originates from the random interference term ΔS(Φ) for each of the components (A;B;C;D) (S40);

- continual updating of the part (S_Translent = ∑ΔS(ID)) of the total damage S that originates from transient irregularities ΔS(ID) for each of the components

(A_JB,C,D) (S41); and

- summation of the total damage S by addition of the part

("S ' _Rmdom = jΔS(Φ)cfr) of the total damage S that originates from the random interference term ΔS(Φ) and the part (S_Translent = £ΔS(ID)) of the total damage S that originates from transient irregularities ΔS(ID) for each of the components.

The invention is not limited to the embodiments described above and illustrated in the drawings, but can be varied within the framework of the following claims. For example, the number of components upon which the estimation is based can be selected in a suitable way. The estimation can be carried out for individual components that are then aggregated using weightings, or for groups of components.

Claims

1) A method for the life prediction of a plurality of vehicle components comprising the following method steps; estimation of the surface roughness of the road upon which the vehicle is travelling, measurement of the speed of the vehicle and estimation of an individual wear index for each component from the parameters surface roughness and speed of the vehicle.

2) The method for life prediction as claimed in claim 1 , characterized in that the parameter surface roughness of the road upon which the vehicle is travelling is determined by means of information about the speed of the vehicle and the vertical acceleration of a component attached to the vehicle.

3) The method for life prediction as claimed in claim 1 or 2, characterized in that the surface roughness of the road upon which the vehicle is travelling is made up of a random interference term and discrete transient irregularities.

4) The method for life prediction as claimed in claim 3, characterized in that the said random interference term is determined from a transfer function between measured vertical acceleration and the road surface roughness.

5) The method for life prediction as claimed in claim 4, characterized in that the said transfer function is determined for a vehicle by means of a derived model of the vertical acceleration with a known road profile as input data. 6) The method for life prediction as claimed in claim 5, characterized in that the model of the vertical acceleration with a known road profile as input data is calculated from a model of the vehicle.

7) The method for life prediction as claimed in any one of claims 3 - 6, characterized in that the random interference term is described as a distribution function in accordance with Θ(Ω) = Θ_OΩ^"W, where Θ_o is the state of the roadway when Ω = 1 rad/m and w = the gradient of the distribution function when Ω

= 1 rad/m.

8) The method for life prediction as claimed in any one of claims 3 - 7, characterized in that discrete transient irregularities are identified by means of a filtering of an input signal from an accelerometer measuring the vertical acceleration.

9) The method for life prediction as claimed in claim 8, characterized in that the said filtered signal is compared with a set of signal responses from a set of known modelled transient irregularities, whereby discrete transient irregularities are approximated with known modelled transient irregularities on the basis of the said filtered signal.

10) The method for life prediction as claimed in any one of claims

8 - 9, characterized in that filtering is carried out by means of a wavelet function.

11 ) The method for life prediction as claimed in any one of the preceding claims, characterized in that an accumulated wear index is created for one or more components by means of a function adapted for each component, comprising a time integration of the size of the random interference term and a summation of discrete transient irregularities.

12) The method for life prediction as claimed in claim 11 , characterized in that the said accumulated wear index for each component is compared with a limit value for the wear index, after which an error message is generated if the wear index exceeds a limit value determined for the component.