Method for detecting friction change in supported shaft of technical system

Abstract

The invention relates to a method for detecting a friction change in a supported shaft (126) of a technical system (100), in which a drive side and an output side of the technical system are connected by means of the shaft, and in which the shaft carries out a rotational movement during operation of the technical system, the method comprises: providing (200) operating variable information (202) relating to one or more operating variables on the drive side of the technical system during operation of the technical system; determining (210) a target value (212) of the rotational speed of the shaft during operation of the technical system on the basis of the operating variable information; determining (220) or providing an actual value (222) of the rotational speed of the shaft during operation of the technical system; determining (230) a difference (232) between the target value and the actual value of the rotational speed of the shaft; and determining (240), based on the difference, whether a friction change (242) of the shaft is present and / or to what extent the friction change of the shaft is present.

Description

Technical Field

[0001] The present invention relates to a method for identifying frictional changes in a supported shaft of a technical system, wherein the drive side and output side of the technical system are connected by the shaft, and to a computing unit and computer program for performing the method. Background Technology

[0002] Turbochargers, also known as exhaust gas turbochargers (ATLs) or, colloquially, turbines, are used to increase the power or efficiency of internal combustion engines. In this case, the turbine impeller and compressor impeller shafts of the exhaust gas turbocharger can be supported hydrodynamically. A device for automatically lubricating rolling bearings is known from WO 01 / 25678 A1. Summary of the Invention

[0003] According to the present invention, a method for identifying frictional variations in a supported shaft of a technical system, as well as a computing unit and computer program for performing the method, are provided, having the features of the independent claims. Advantageous designs are the subject of the dependent claims and the following description.

[0004] This invention studies the shaft of a technical system, such as the shaft of an exhaust gas turbocharger of an internal combustion engine, which rotates during the operation of the technical system, wherein the shaft carries a turbine impeller or a compressor impeller. Here, the turbine impeller is arranged on the drive side, while the compressor impeller is arranged on the output side of the turbocharger.

[0005] An exhaust gas turbocharger (ATL) is an auxiliary unit in an internal combustion engine used to compress the supplied air. A higher mass flow of air and therefore a larger quantity of oxygen allows the internal combustion engine to operate more efficiently, that is, to increase engine power. A turbocharger in its simple form consists of a turbine impeller driven by the energy of the exhaust gas and a compressor impeller that uses the driving energy to increase the air pressure in the intake system on the output side. The turbine impeller and compressor impeller are typically anti-rotatingly fixed to a shaft and hydrodynamically slidably supported in a housing.

[0006] Other components of a turbocharger include a boost air cooler and an exhaust valve system, which are used for boost regulation and to prevent ATL overload. In addition to the turbochargers mentioned above, there are other types of turbochargers, such as turbochargers with variable turbine geometry (VTG).

[0007] Because ATL shafts typically experience very high rotational speeds (200,000 U / min is not uncommon), hydrodynamic sliding bearing structures bear particularly strong loads. If the lubricating film breaks down and the resulting change in the friction ratio, especially solid friction, this can lead to shaft seizure and breakage within a short period of time (usually within seconds to minutes).

[0008] Against this backdrop, a feasible solution is proposed to identify such damage conditions, namely frictional changes, in advance and to prevent complete failure.

[0009] The proposed method will be illustrated in particular with the example of an exhaust gas turbocharger; however, the method can also be applied to other shafts of a technical system that perform rotational motion during operation.

[0010] Here, operating parameter information regarding one or more operating parameters of the drive side of the technical system is provided during operation. These operating parameters of the drive side of the technical system may include, for example, exhaust gas temperature, exhaust gas mass flow, and exhaust gas pressure. Additional parameters affecting drive power and friction power, such as oil temperature and exhaust valve mass flow, are also considered.

[0011] The target value of the shaft rotation speed is then determined based on operational parameter information during the operation of the technical system. For this purpose, a machine learning model is used in an appropriate manner, as will be explained in more detail below.

[0012] Furthermore, the actual value of the shaft's rotational speed is determined or provided during the operation of the technical system. To this end, in one embodiment, an acoustic signal is provided, wherein the acoustic signal is detected from the technical system or another technical system mechanically connected to the technical system by means of a sensor (e.g., a solid-state acoustic sensor) during the operation of the technical system. The actual rotational speed is then determined based on the acoustic signal.

[0013] To this end, for example, the time signal is transformed into the frequency domain, that is, the spectral power density is obtained (e.g., using the Welch method). In order to obtain the characteristic speed of the exhaust gas turbocharger, the dominant tone component (e.g., 1st order, 6th order, etc. in the case of six compressor impellers) can be found, for example, by a lookup function (e.g., Peak-Find implemented in the Python library scipy).

[0014] However, in one implementation, if a speed sensor is present, the actual speed is also determined based on the measurements taken by that speed sensor.

[0015] Furthermore, the difference between the target and actual values ​​of the shaft rotational speed is then determined, and based on this difference, it is determined whether and / or to what extent the shaft friction variation exists. Specifically, if the difference exceeds a pre-defined threshold (e.g., a critical measure), the presence of shaft friction variation is determined. The extent of shaft friction variation can also be determined based on the time variation of the difference, for example.

[0016] The proposed method is based on the assumption that, under varying friction ratios, the same drive power of a turbocharger (or technical system) will result in a change in friction power. If the friction power increases, this will lead to a decrease in the actual value of the measured or observable rotational frequency (and therefore the shaft speed) under the same drive power. The physical principle underlying this consideration is the shift from viscous friction to solid friction, which dominates in normal operating characteristics. It should be noted here that rotational frequency and rotational speed can be understood essentially synonymously.

[0017] Typically, the so-called Stribeck curve is considered in conjunction with the relative velocity between the two friction pairs (an example of a Stribeck curve is shown in the attached figure). However, in addition to the relative velocity, the viscosity of the lubricating film and the normal force are also input into the function of the coefficient of friction.

[0018] Poor lubrication (insufficient oil or fluid) in the bearing components can lead to a thinning of the oil film on the bearing shaft, and a shift in friction from viscous friction (hydrodynamic lubrication) to mixed friction or solid friction. This can result in a decrease or increase in frictional power within the bearing. Here, mixed friction can mean, for example, a decrease in friction, or a shift to solid friction can mean, for example, an increase in friction. However, some Stribek curves exhibit a parabolic trend in the mixed friction range, meaning that depending on the rotational speed, there may be both increases and decreases in friction within the mixed friction range. Furthermore, reduced oil supply can cause changes in kinematics and thus generate localized temperature peaks, which in turn lead to localized decreases in viscosity.

[0019] Therefore, it is possible to identify potential frictional changes based on frequency variations (i.e., the difference between the target and actual values), and thus anticipate impending (but possibly yet to occur) damage. Similarly, it is possible to infer, for example, the use of different or incorrect fluids or oils with varying viscosities based on frictional changes.

[0020] Here, for example, a machine learning model is used to determine the target value of the rotational speed, which corresponds to a common operating parameter on the drive side under normal or good conditions. This operating parameter, or its value, does not change with friction variations; however, the actual value of the rotational speed changes, leading to the aforementioned difference.

[0021] To identify such damage in advance and to prevent complete failure, the changing friction ratio can be identified using, for example, auxiliary parameters from the air system and, if necessary, solid-borne acoustic signals measured at the turbocharger housing. This can largely prevent unplanned downtime and costly repairs or indirect damage.

[0022] Generally, this provides various advantages for hydrodynamic sliding bearing structures, and especially for ATL-shaft bearing structures. It allows the detection of varying friction ratios, which may be attributed to reasons such as: radial misalignment of the shaft due to incorrect installation (e.g., so-called skewing), shaft eccentricity, deviation from tolerance machining dimensions, incorrect balancing, etc. Also, there are variations in the bearing surface of the shaft due to disturbed pressure distribution, insufficient oil supply due to blocked pipes. Furthermore, there are the installation or addition of incorrect bearing sleeves (material properties, such as surface roughness, can also affect friction characteristics), oil contamination (particles / wear can also alter friction), or the use of incorrect oil (especially incorrect viscosity).

[0023] The advantage also lies in the ability to characterize friction using easily measurable solid-borne sound and parameters continuously detected in the air system, without the need for expensive, complex sensing devices to detect oil properties. Since the method only requires measurements in the turbocharger (or technical system) in good condition and can be trained specifically for the turbocharger, it can also be used as a retrofit solution.

[0024] The computing unit according to the invention, such as a controller of a motor vehicle or a computer, is particularly configured in terms of programming technology to execute the method according to the invention.

[0025] The implementation of the method according to the invention in the form of a computer program or computer program product having program code for executing all method steps is also advantageous, as this results in particularly low costs, especially if the implemented controller is also used for other purposes and therefore already exists. Finally, a machine-readable storage medium is provided having the computer program as described above stored thereon. Suitable storage media or data carriers for providing the computer program are, in particular, magnetic, optical, and electrical memories, such as, for example, hard disks, flash memory, EEPROM, DVDs, etc. The program can also be downloaded via a computer network (Internet, intranet, etc.). Here, such downloading can be achieved wired or wirelessly (e.g., via WLAN networks, 3G, 4G, 5G, or 6G connections, etc.). Attached Figure Description

[0026] Further advantages and design solutions of the present invention are available from the specification and drawings.

[0027] The present invention is schematically illustrated with reference to the embodiments shown in the accompanying drawings and is described below with reference to the drawings. Wherein: Figure 1 The technical system in which the present invention can be used is illustrated schematically; Figure 2 The flow of the method is illustrated in one implementation; Figure 3 and Figure 4 A diagram is shown to illustrate the invention. Detailed Implementation

[0028] exist Figure 1 The diagram schematically illustrates a technical system 100 configured as an exhaust gas turbocharger, in which the present invention can be used. A cross-section of an internal combustion engine 1 having a cylinder or cylinder block 150 and an exhaust gas turbocharger 100 is schematically shown here. The cylinder block 150 has an air inlet 151 for supplying fresh air and an exhaust bend 152 for discharging exhaust gases.

[0029] The exhaust gas turbocharger 100 has a turbine 120 on the drive side and a compressor 110 driven by the turbine on the drive side. Fresh air 10 is supplied to the compressor 110, which compresses it to the boost pressure and supplies it to the cylinder block 150 via an optional air cooler 11. In the cylinder block 150, a fuel / air mixture is known to be combusted, and the resulting exhaust gas is supplied to the turbine 120 via an exhaust bend 152.

[0030] The turbine 120 has a turbine housing 123 with a turbine inlet 121 and a turbine outlet 122, and between them a guide vane 124 with adjustable guide vanes and a turbine impeller 125. Exhaust gas enters the turbine housing 123 under pressure at the turbine inlet and flows through the turbine housing 123 and the guide vane 124, where the pressure is throttled at a transfer section between the guide vane and the turbine impeller. The exhaust gas drives the turbine impeller 125, and then flows out at the turbine outlet 122 into the exhaust system 12.

[0031] Here, the turbine impeller 125 is coupled to the compressor impeller 115 of the compressor 110 via a shaft 126. During operation, the turbine impeller 125 and the compressor impeller 115 rotate at a speed denoted here by n. Here, the shaft 126 is hydrodynamically slidably supported.

[0032] The exhaust gas turbocharger shown here is merely an example; thus, for example, an exhaust gas turbocharger with an exhaust valve can also be used.

[0033] Furthermore, a computing unit 170 configured as a motor controller and a solid-state acoustic sensor or knock sensor 160 disposed on the housing of the exhaust gas turbocharger 100 are shown. For example, it is also conceivable that the solid-state acoustic sensor or knock sensor 160 is disposed on the cylinder block 150. In addition, a speed sensor 162 is provided by way of means by which the rotational speed of the exhaust gas turbocharger or the shaft 126 of the exhaust gas turbocharger can be detected.

[0034] As mentioned, damage can occur in such exhaust gas turbochargers due to frictional variations caused, for example, by poor lubrication of the hydrodynamic support structure. The identification of such frictional variations will be explained in more detail below.

[0035] exist Figure 2 The flow of the method is illustrated in one embodiment.

[0036] In step 200, operating parameter information 202 regarding one or more operating parameters is then provided during operation. This relates to operating parameters on the drive side, specifically on the side of the turbine impeller. In step 210, a target value 212 for the rotational speed of the exhaust gas turbocharger shaft is determined based on this.

[0037] The target value (target RPM) can then be obtained from the ATL's input parameters, namely exhaust gas temperature, exhaust gas mass flow, exhaust gas pressure, and possibly additional parameters affecting drive power and friction power. Additional parameters include, for example, oil temperature and exhaust valve mass flow, or other parameters mentioned below. It is important to note which parameters from the air system are used as independent variables. Other methods typically incorporate primarily parameters from the compressor side of the turbocharger, while the proposed method specifically uses only drive parameters (turbine side).

[0038] The aforementioned operating parameters can be detected using suitable sensors. For example, thermocouples / resistance thermometers can be considered for exhaust gas temperature; hot-film air volume meters / differential pressure meters, or modeling / virtual sensing devices using other controller parameters such as motor torque and motor speed, can be considered for exhaust gas mass flow; thermocouples or thermistors (semiconductor resistors) can be considered for oil temperature; changes in resistance, capacitance, or voltage when mechanical pressure is applied to a diaphragm can be considered for exhaust valve position; potentiometers or Hall sensors can be considered for exhaust valve mass flow; fresh air temperature can be considered for exhaust gas temperature; the temperature after the turbine can be considered for exhaust gas temperature; and the pressure after the turbine can be considered for exhaust gas pressure.

[0039] To detect parameters, it is typically possible to use, for example, sensors that are installed in batches in motor vehicles.

[0040] For simplicity, we can assume here that the exhaust gas pressure and temperature at the exhaust valve and turbine inlet are the same. The data-driven model, that is, the machine learning model, can learn about independent variables, ATL-speed, from the relationships between the mentioned operating parameters. or rotation frequency The mapping (e.g., as a so-called regression model).

[0041] The mapping rule for the target speed is learned in a suitable manner under optimal turbocharger conditions, that is, under conditions of optimal friction ratio. Therefore, the mapping rule is formally applicable: Here, for example, To represent the exhaust gas mass flow and in This is used to represent exhaust gas pressure.

[0042] Different methods (such as linear regression, Gaussian processes, neural networks, etc.) can be used to create machine learning or regression models. It's important to note that turbine-side or drive-side parameters are used as input data for the model because the frictional torque in the support structure reacts to the drive torque, and the compressor power is reflected in steady-state operation.

[0043] Furthermore, in step 220, the actual value 222 of the shaft rotational speed is determined or provided. This can be achieved by using the aforementioned speed sensor or by providing an audible signal 226 (step 224).

[0044] For training to learn mapping rules, there should also be observed characteristic rotational frequencies from sound signals (e.g., airborne or solid-borne sounds) as labels. As already mentioned, a feasible solution hereafter is an additional speed sensor at the ATL. Since the turbine impeller and compressor impeller are anti-rotationally connected to each other on the shaft, they both have the same rotational speed.

[0045] However, the availability of speed sensors at the ATL is limited in mass production. Therefore, another feasible approach for training is also based on the acquisition of sound or solid-state acoustic signals. This allows for the transformation of the time signal (i.e., the sound signal existing in the time domain) into the frequency domain, that is, the acquisition of spectral power density and / or amplitude spectrum, for example, using the Welch method. Spectral information or a spectrogram is then obtained. Specifically, for example, a spectral representation is obtained for each time interval, where the spectrogram is a visual representation of the signal's spectral information over time. For example, the spectral power density is plotted on the time axis (horizontal) and the frequency axis (vertical). Color levels / brightness represent the signal intensity within that range.

[0046] Here, frequency domain transformation is specifically understood as a signal transformation that converts a signal from the time domain to the frequency domain. For this purpose, variations of the Fourier transform can be used, for example (typically the Discrete Fourier Transform, Fast Fourier Transform, etc., used for digital signals). Here, the signal is decomposed into its frequency, amplitude, and phase components.

[0047] Therefore, in Figure 3 The spectrum diagram 300 is shown as an example, with frequency 302 plotted with respect to time 304. There, the actual value of the rotational speed is represented by 310, and higher or lower values ​​representing changes in friction are represented by 312 and 314.

[0048] In order to obtain the characteristic speed of the ATL, in this case, the actual value, the principal component can be found using a lookup function (such as Peak-Find). This principal component can be, for example, first order or order related to the number of compressor impellers, such as sixth order in the case of six compressor impellers.

[0049] Besides visual observation of the spectrum, there are feasible methods for estimating the ATL speed. In this regard, "virtual sensing devices" can also be mentioned. This can be done, for example, using a machine learning model that takes the operating parameters of the TL or a typical technical system as input data. These operating parameters are those that differ from the ATL speed, particularly thermodynamic parameters such as air mass flow and pressure ratios in the exhaust gas turbocharger.

[0050] If the actual rotational speed is not continuously visible in the spectrum, it is possible to perform additional filtering functions on the operating parameters, for example.

[0051] Model training can be included within a preprocessing framework, encompassing common steps (if needed) when creating models using machine learning methods, such as removing redundant parameters via correlation filters, data cleaning, or data standardization. However, in addition to data-driven models (machine learning models), physical modeling can also be envisioned—that is, using a physical model that can determine the actual rotational speed (the actual value of the rotational speed) based on runtime parameter information. Here, it is necessary to consider the necessary physical relationships.

[0052] In step 230, the difference 232 between the target value 212 and the actual value 222 of the shaft rotational speed is then determined. Based on the difference, in step 240, it is then determined whether there is a change in shaft friction 242 and / or the extent to which the change in shaft friction exists.

[0053] The determination of the difference corresponds to the observation of the residual, and more specifically, to the testing of the condition: .

[0054] If the observed actual rotational speed from the spectrum or spectrogram corresponds to the model prediction of the regression model, then the difference is less than the critical measure. (i.e., the threshold) and the condition is not met. This means that the friction ratio has not changed compared to the training process, i.e., compared to the good state.

[0055] In addition to observing instantaneous values, the frequency distribution of residuals over a certain time interval can also be considered. Therefore, clustering multiple values ​​can improve robustness and, if necessary, account for changes in the operating point. Friction variations can also be described by changes in the shape of the distribution (e.g., Kolmogorov-Smirnov tests, central moments, etc.).

[0056] Residual observation thus obtains, for example, the difference (or residual) between the actually observed frequency and the expected frequency (calculated by the model). Here, the model learns this relationship under ideal conditions for the turbocharger. If the residual exceeds a certain threshold, this can indicate an increase or decrease in friction, that is, an error exists in the system.

[0057] Conversely, if the critical measure An increase indicates a change in the friction ratio. According to... Figure 3 This also clarifies the point. A rotational speed or frequency of 310 corresponds to a normal or good state, that is, to the target rotational speed. However, with changes in friction, the rotational speed, that is, the actual value, can shift upwards or downwards, that is, towards 312 or 314. This results in a difference. It should be noted that as friction increases, the frequency or rotational speed decreases, and as friction decreases, the frequency or rotational speed increases.

[0058] In addition to the absolute value, dynamic characteristics may also change due to varying friction, such as a slower start or a more severe delay when the shaft is running inertia under no-load conditions.

[0059] As already mentioned, the proposed approach is based on the assumption that the same drive power of a turbocharger, coupled with a varying friction ratio, will result in a change in frictional power. If the frictional power increases, this, under the same drive power, will lead to a decrease in the measured or observable rotational frequency and, consequently, the rotational speed of the shaft. The physical principle upon which this consideration is based is the shift from viscous friction, which dominates in normal operating characteristics, to solid friction.

[0060] exist Figure 4 This is illustrated by the so-called Stribek curve 400. The coefficient of friction 402 is plotted there, and more precisely, it is shown with respect to relative velocity, lubricating film viscosity, and normal force along with 404.

[0061] Typically, the Stribek curve is considered in conjunction with the relative velocity between the two friction pairs. However, as mentioned here, in addition to relative velocity, the viscosity of the lubricating film and the normal force are also input into the function of the coefficient of friction.

[0062] Poor lubrication (insufficient oil or fluid) in the support components can cause the oil film on the bearing shaft to thin, and friction to... Figure 4 The viscous friction (hydrodynamic lubrication) shown in 410 shifts to the range of mixed friction shown in 412 or solid friction shown in 414. As mentioned, the Stribek curve can also behave differently.

[0063] For example, poor lubrication occurs when the input line of a hydrodynamic sliding support structure is throttled due to damage or contamination.

[0064] Furthermore, reduced oil or fluid supply can alter kinematic relationships and generate localized temperature spikes, which in turn lead to localized viscosity reductions. This can potentially reduce frictional resistance to the point that damage identification must be based on the residual (ATL's target speed minus ATL's observed actual speed) in terms of value.

[0065] In summary, it can be determined that, unlike other feasible methods that react to changes in amplitude to identify damage, the proposed method is based on the observation of frequency shifts, specifically the difference between the actual and target values ​​of the rotational speed (or rotational frequency). In particular, the proposed method focuses on how the tonal components of the signal generated by the rotational speed of the drive (ATL or other technical system's drive side) shift to lower or higher frequencies while the drive power remains constant. This effect is caused by increased or decreased friction, which acts as a marker for potential errors.

[0066] In this way, damage that has not yet caused frequency-specific amplitude changes in the sound signal can be detected. For the proposed method, there is no need for (obvious) force excitation within the machine (e.g., due to imbalance, pitting at rolling bearings) that would cause a change in the sound signal, specifically the amplitude. A specific frequency amplitude change is not mandatory. It is sufficient that the frictional characteristics change—which can also be caused, for example, by other oils with different viscosities—and that the frequency change can be observed in the actual rotational speed parameters under the same or similar drive parameters. Therefore, using the proposed method, (potential future) damage can be identified at an earlier stage.

[0067] Furthermore, the proposed method does not necessarily require the use of a solid-state acoustic sensor or receiver; a rotational speed receiver can also be used to measure the shaft rotational speed. Traditionally, shaft rotational speeds are compared at the same operating point. However, by using a machine learning model, the correlation between operating points can be directly mapped, eliminating the need for a reference operating point and, in principle, enabling the detection of errors at each operating point where frictional changes are observed.

[0068] Therefore, it is particularly capable of monitoring the condition of hydrodynamic sliding bearings, that is, by monitoring frictional changes during the operating duration of the technical system.

[0069] The proposed method can be used not only in field applications for online diagnostics but also for inspection in manufacturing processes (e.g., end-of-line inspection). Bandpass filtering is feasible. In addition to the first rotational speed order mentioned, other rotational speed orders or combinations thereof that provide better visualization can also be used. For acoustic signals, acceleration, velocity, or position at a surface can be measured.

Claims

1. A method for identifying frictional variations in a supported shaft (126) of a technical system (100), wherein, The method involves connecting the drive side and the output side of the technical system via the shaft, wherein the shaft performs rotational motion during operation of the technical system. During the operation of the technical system, provide (200) operational parameter information (202) about one or more operational parameters of the driving side of the technical system; During the operation of the technical system, a target value (212) of the rotational speed of the shaft is determined (210) based on the operating parameter information. During operation of the technical system, determine (220) or provide the actual value (222) of the rotational speed of the shaft. Determine (232) the difference between the target value and the actual value of the rotational speed of the shaft (230); And based on the difference, determine (240) whether there is a frictional change in the shaft (242) and / or to what extent the frictional change in the shaft exists.

2. The method according to claim 1, wherein, If the difference is greater than a pre-given threshold, then a frictional change in the shaft is determined to exist.

3. The method according to claim 1 or 2, wherein, The extent to which frictional changes in the shaft exist is determined based on the time variation of the difference.

4. The method according to any one of the preceding claims, wherein, The target value of the rotational speed of the shaft is determined using a machine learning model (214).

5. The method according to any one of the preceding claims, further comprising: Provide (224) an audio signal (226), wherein the audio signal is detected by means of a sensor (160) from the technical system (100) or another technical system mechanically connected to the technical system during operation of the technical system; The actual rotational speed is determined based on the sound signal.

6. The method according to claim 5, wherein, Solid-state acoustic sensors (160), especially knock sensors, are used as sensors.

7. The method according to any one of the preceding claims, wherein, The actual rotational speed is determined based on the measurements taken by the rotational speed sensor (162).

8. The method according to any one of the preceding claims, wherein, One or more operating parameters on the drive side of the technical system include at least one of the following operating parameters: exhaust gas temperature, exhaust gas mass flow, exhaust gas pressure, oil temperature, and exhaust valve mass flow.

9. The method according to any one of the preceding claims, wherein, Exhaust gas turbochargers are used as a technical system.

10. A computing unit configured to perform all method steps of the method according to any one of the preceding claims.

11. A computer program, when implemented on a computing unit, causes the computing unit to perform all the method steps of the method according to any one of claims 1 to 9.

12. A machine-readable storage medium having a computer program stored thereon according to claim 11.

Images