DE102021119967A1 - METHOD AND SYSTEM FOR NON-CONTACT, NON-DESTRUCTIVE REAL-TIME COMPONENT MONITORING - Google Patents

Abstract

The invention relates to a method and a system for non-contact, non-destructive real-time component monitoring. First, a calibration system is provided with a first component and a first support and guide element, the first component not having a crack and/or damage and being in contact with the first support and guide element. A transfer path analysis is then carried out on the calibration system. In a further method step, a second component is moved along a longitudinal axis of a second carrying and guiding element, the second component being in continuous contact with the second carrying and guiding element. Furthermore, vibrations in the second support and guide element are recorded by at least one detection device attached to the second support and guide element. These recorded vibrations are transformed using a transfer path and synthesis model, so that an extracted vibration spectrum of the second component is obtained. In a final method step, the vibration spectrum of the second component is analyzed, with the analysis being able to detect a crack and/or damage.

Description

The invention relates to a method and a system for non-contact, non-destructive real-time component monitoring. In a first method step, a calibration system is provided. This calibration system comprises a first component and a first support and guide element, the first component not having a crack and/or damage and being in contact with the first support and guide element. A transfer path analysis is then carried out on the calibration system. In this case, transfer functions of transfer paths are determined and a transfer path model is generated based on the transfer paths involved. This transfer path model is verified by a microphone measurement. Furthermore, a synthesis model is generated by inverting the transfer path model. In a further method step, a second component is moved along a longitudinal axis of a second carrying and guiding element, the second component being in continuous contact with the second carrying and guiding element. Furthermore, vibrations in the second support and guide element are recorded by at least one detection device attached to the second support and guide element. These recorded vibrations are transformed using the transfer path and synthesis model, so that an extracted vibration spectrum of the second component is obtained. In a final method step, the vibration spectrum of the second component is analyzed, with the analysis being carried out with regard to natural frequencies, damping and/or an amplitude, and as a result a crack and/or damage can be detected.

Background and prior art

Wheel tires of rail vehicles are subject to considerable wear due to high stress. Knowing the state of wear of the wheels is essential both for safe railway operation and for avoiding accidents with extreme property damage and personal injury.

• • Gleisbelegungszeit - die Zeit, während ein Fahrzeug mit Radfehlern zurückgelassen werden muss.
• • schnellere Degradierung des Gleises - mit großen dynamischen Kräften durch defekte/unrunde Räder, die zu einer früheren und intensiveren Instandhaltung des Gleises führen.
• • Geschwindigkeitsbeschränkungen - während ein Fahrzeug mit einem defekten Rad in eine Werkstatt gebracht wird und bevor eine Inspektion des Gleises stattgefunden hat

Damage to a railway vehicle wheel has fatal effects on the operating capacity due to:

• • Track Occupancy Time - the time during which a vehicle with wheel faults must be left behind.
• • Faster track degradation - with large dynamic forces from defective/out-of-round wheels leading to earlier and more intensive track maintenance.
• • Speed limits - while a vehicle with a defective wheel is brought into a workshop and before a track inspection has taken place

Typical damage to wheel tires includes flat spots, cracks, fractures and loosening of the interference fit on the wheel disc. Initially, small hairline cracks are not an immediate danger for a sudden tire failure. They are often side effects of the application process to the wheel disc due to shrinkage. On the other hand, loosening of the wheel tire on the wheel disc is a major source of danger and can lead to serious accidents with derailments. Loose wheel tires must be recognized and eliminated immediately. Waiting for the next inspection interval may already be too late.

Due to their safety relevance, wheel tire systems often have to be checked in a workshop for routine inspections. However, due to the necessary shutdown of entire trains, conventional condition monitoring methods are extremely time-consuming and resource-consuming. This leads to a reduction in traffic capacity, combined with an increase in train delays and cancellations. The optimization of the detection process with regard to component damage on the wheels of rail vehicles is therefore of great importance for the timely replacement of the worn component or a repair. In view of the risks of railway operations, the monitoring of technical operational safety has always been preventive. Components and systems are monitored at prescribed maintenance intervals.

A large number of methods are known for detecting damage to rail vehicle wheels. There are "in workshop" test methods and "in service" test methods for detecting defects.

A major disadvantage of the "in workshop" test procedure is the need to dismantle individual wheelsets from the train. Due to the necessary decommissioning of entire trains, such condition monitoring methods are extremely time-consuming and resource-consuming. This leads to significant maintenance costs, a reduction in traffic capacity with a consequent increase in train delays and cancellations.

"In service" measurement methods are on-board measurement methods (ultrasonic, magnetic, airborne and structure-borne) and trackside (strain gauges, fiber optic sensor technology, ultrasonic, structure-borne, airborne, and laser and high-speed cameras).

The main disadvantage of the "in operation" test method is in particular the measurement inaccuracy and susceptibility to errors due to the robust and complex environmental conditions. The measurement of acoustic states in vehicle operation and the automated determination of component states are subject to significant limitations due to background noise and environmental influences. Acoustic monitoring systems use sound level meters or microphones and axle counters. Vehicle monitoring in particular uses additional sensors and technologies to determine the condition of a vehicle.

From the pamphlet U.S. 2016/0207552 A1 a system and a method for detecting flat spots on track-bound wheels while driving is known. Acceleration sensors for picking up vibrations are disclosed, these being placed at a distance from one another along a rail arrangement. The recorded vibrations generate a sound image, which is analyzed by examining the sound image for conspicuous vibration deflections. However, the disadvantage of the method and system is that it is only set up to detect flat spots on a wheel and is also not suitable for detecting further damage or for classifying the type of damage. Furthermore, the system or method is very imprecise and error-prone, since the measurements are also subject to the influence of other components such as the rail arrangement, the carriage axle or other components.

The pamphlet DE 10 2018 123 445 A1 relates to the identification of noise sources, in particular freight wagons and trains, for legal protection without background noise. The teaching disclosed aims in particular at determining the volume of a train without interference noise such as traffic noise (car, helicopter) affecting a volume result. Particularly preferably, the volume of each individual track-bound wheel can be determined. However, the disclosed method or system does not enable damage detection and in particular no detection of the type of damage (e.g. crack, loosening or flat spot).

In addition to checking component damage on trains, especially rail-bound wheels, quality assurance in the production of mass-produced items, for example, is still of great relevance. Due to the large number of articles to be manufactured, only random samples are often selected for a suitable quality check and examined for damage. Such a quality check is very complex and is often still carried out manually by assessment by a trained specialist. There is therefore a need for an automatic, non-contact, non-destructive and early detection of damage to components, particularly in mass production.

object of the invention

The object of the invention was therefore to eliminate the disadvantages of the prior art and to provide a robust and error-resistant method and system for non-contact, non-destructive real-time component monitoring, with damage as such and the type of damage being identified.

Summary of the Invention

The object according to the invention is solved by the features of the independent claims. Advantageous configurations of the invention are described in the dependent claims

1. a. Bereitstellung eines Kalibriersystems, welches ein erstes Bauteil und ein erstes Trag- und Führungselement umfasst, wobei das erste Bauteil nicht mit einem Riss und/oder Schaden behaftet ist und in Kontakt mit dem ersten Trag- und Führungselement steht;
2. b. Durchführen einer Transferpfadanalyse am Kalibriersystem, wobei
   • - Übertragungsfunktionen von Transferpfaden ermittelt werden;
   • - ausgehend von den beteiligten Transferpfaden ein Transferpfadmodell erzeugt wird;
   • - das Transferpfadmodell durch eine Mikrofon-Messung verifiziert wird;
   • - ein Synthesemodell erzeugt wird, indem das Transferpfadmodell umgekehrt wird;
3. c. Bewegen eines zweiten Bauteils entlang einer Längsachse eines zweiten Trag- und Führungselementes, wobei das zweite Bauteil durchgehend in Kontakt mit dem zweiten Trag- und Führungselement steht;
4. d. Erfassen von Schwingungen im zweiten Trag- und Führungselement durch mindestens eine am zweiten Trag- und Führungselement befestigte Erfassungseinrichtung;
5. e. Transformieren der erfassten Schwingungen mittels des Transferpfad- und Synthesemodells, sodass ein extrahiertes Schwingungsspektrum des zweiten Bauteils erhalten wird;
6. f. Analyse des Schwingungsspektrums des zweiten Bauteils, wobei die Analyse im Hinblick auf Eigenfrequenzen, Dämpfung und/oder eine Amplitude erfolgt und daraus resultierend ein Riss und/oder ein Schaden erkannt werden kann.

In a preferred embodiment, the invention relates to a method for non-contact, non-destructive real-time component monitoring, characterized in that the method comprises the following steps:

1. a. Providing a calibration system, which comprises a first component and a first support and guide element, wherein the first component is not afflicted with a crack and/or damage and is in contact with the first support and guide element;
2. b. Carrying out a transfer path analysis on the calibration system, wherein
   • - Transfer functions of transfer paths are determined;
   • - Based on the transfer paths involved, a transfer path model is generated;
   • - the transfer path model is verified by a microphone measurement;
   • - a synthesis model is generated by inverting the transfer path model;
3. c. moving a second component along a longitudinal axis of a second support and guide member, the second component being in continuous contact with the second support and guide member;
4. i.e. Detection of vibrations in the second support and guide element by at least one detection device attached to the second support and guide element;
5. e. Transforming the detected vibrations using the transfer path and synthesis model, so that an extracted vibration spectrum of the second component is obtained;
6. f. Analysis of the vibration spectrum of the second component, with the analysis being carried out with regard to natural frequencies, damping and/or an amplitude and, as a result, a crack and/or damage being able to be detected.

The starting point for the present method was the knowledge that the condition (in terms of damage) of a component can be determined by sound, with the structure of the component being caused to vibrate by an impulse excitation (e.g. force impulse from a hammer blow). Due to its structural nature, the component is more sensitive to certain frequencies, which means that these frequencies are more pronounced than others. Since such an analysis requires excitation by a pulse, non-contact component monitoring using conventional methods known from the prior art cannot be implemented.

The core of the method according to the invention is to use an acoustic resonance analysis to arrive at a precise diagnosis of the condition of a component—without subjecting the component to active impulse excitation (or touching). This has presented itself as a great challenge and was not possible without inventive considerations. Observations have shown that the movement of a component that is in contact with a support and guide element causes vibrations in the support and guide element. A special (according to the invention) analysis of the vibrations leads to conclusions about the condition of a component, with the analysis being carried out so precisely that not only the mere presence of damage (i.e. an abnormality) but also the type of damage can be recognized.

The combination of the proposed process steps leads to a surprising synergy effect, which leads to the advantageous properties and the associated overall success of the invention, with the individual features interacting with one another. An important advantage of the method according to the invention is the need for extremely few method steps and system components, while an extremely robust and fault-resistant infrastructure for non-contact, non-destructive real-time component monitoring is nevertheless generated. Due to its few system components and method steps, the method can advantageously be implemented in a procedure/process—such as a production process or a train journey—in a particularly simplified manner. Extensive conversions of existing systems, cuts or changes in the sequence of operations/processes are not necessary. On the contrary, the method according to the invention is advantageously set up to run in parallel with a conventional procedure/process without having any effect on this procedure/process. The innovation of the method according to the invention also consists in the fact that damage and/or a crack can be detected fully automatically during an ongoing process/operation, the method also having an extremely high sensitivity. As a result, even minor damage—which may not be visible and/or recognizable in a manual examination by a person—can be determined using the method according to the invention. In addition, the method enables damage to be detected within an extremely short time, namely immediately (preferably within milliseconds) after the vibrations have been recorded.

The proposed method as a solution to the above task is to be regarded as a novel technology of so-called "predictive maintenance", which uses condition and performance data of components to derive maintenance information in the sense of predictive maintenance. The aim is to proactively maintain machines and systems or their components and to minimize downtimes. In the ideal case, faults can be predicted before they have an impact or failures. through early If maintenance measures are taken proactively, the actual occurrence of the fault can be prevented.

The method according to the invention is preferably also to be understood as a method for detecting cracks and/or damage to a component, with a component preferably being understood as an individual part of a technical complex (e.g. machine, apparatus, etc.) within the meaning of the invention. A component can be a fully manufactured object that is ready for use (or is in use), or the component can also be a semi-finished product or a starting material. A component can therefore preferably be installed in a machine (or the like) and monitored while a process/operation is being carried out or while a machine is idle. On the other hand, the component can also be examined for damage during, before or after its manufacture/production. The use of the present method is particularly advantageous when monitoring components that are installed in complex machines, because the method enables component monitoring without having to remove the component from the machine. Above all, the non-contact and non-destructive process ensures that the component is still functional after the inspection.

A support and guide element is preferably to be understood as a component which is set up to support and/or guide a component according to the invention and/or a machine comprising a component. This includes, for example, a workpiece carrier, a conveyor belt system in manufacturing/production, a discharge chute from a production machine for finished components and/or rails for rail-bound machines or vehicles. The support and guide element is not limited to this and can assume all other configurations known to a person skilled in the art.

The component (the first and/or second component) is particularly preferably in contact with the supporting and guiding element (the first and/or second supporting and guiding element), whereas the supporting and guiding element is preferably in contact with a floor (element ) or another component connected to a floor, on which the support and guide element is set up and in which emerging forces are preferentially introduced. This means that there is physical touch or physical contact between the component and the supporting and guiding element and between the supporting and guiding element and a floor (element) or another component connected to a floor.

The proposed method can preferably be divided into two sections, namely a calibration section (method steps a and b) and an examination section (method steps c to f). It goes without saying that a calibration is carried out first and then a component (or a large number of components) can be examined for damage and/or cracks.

A calibration system is preferably provided in the calibration section of the method. The calibration system preferably serves as a reference system. The calibration system includes a first component and a first support and guide element, the first component not having a crack and/or damage and being in contact with the first support and guide element. The calibration of a calibration system makes it possible, in particular, to provide the method according to the invention with a reference to a good part (first component), that is to say a non-damaged component. In addition, the accuracy of the method can be increased and the susceptibility to errors can be minimized by the calibration.

In the course of the calibration, a transfer path analysis is preferably first carried out on the calibration system. A transfer path analysis analyzes in particular the propagation paths (transfer paths) of sound from its point of origin to a measuring point through a medium, which e.g. B. can be a fluid (especially air) or a solid. Multiple media are often involved in the propagation of sound. A transfer path analysis can, for example, be carried out experimentally, in the form of measurements directly at the sound source and at more distant locations in the medium being analyzed. Also, theoretical calculations of transfer paths are based on the properties of the emitted sound such. B. frequency, amplitude and phase as well as the media involved in the transmission possible, but may be complex.

The transfer path analysis is also preferably to be regarded as an analysis of the transfer properties of an oscillating system (here: preferably a calibration system). A transfer path describes the transmission of a signal, it being possible for the transfer path to be represented by one or more transfer functions. The calibration system preferably transforms an excitation signal into a receiver signal, where this transformation can be represented by one or more transfer function(s) H (ω) (cf. also 7 ).

The transfer functions of the transfer paths of the airborne and structure-borne noise from the noise source to the measurement location should preferably be determined. Airborne sound preferably refers to the direct propagation of the sound from the sound source to the preferred measurement location, so the medium in this case is in particular air. Above all, this can refer to the direct sound propagation between the sound source and the measurement location along the "linear distance" (the direct connection). However, sound reflected from other media can also be part of the airborne sound. It is primarily about the sound, which only reaches the measuring point through air and not through another medium.

Structure-borne noise preferably means the indirect propagation of the noise from the noise source to the measurement location through the (solid) bodies connected to the noise source. There can also be a partial propagation of the sound from the body involved in the propagation of the sound to the measuring point through the air. Structure-borne noise primarily refers to the part of the noise that occurs at the contact between the component and the support and guide element and the part of the sound that occurs at the contact between the support and guide element and a floor (or another component connected to a floor), which is at least partially also through a medium other than air, in particular through a solid body. If one considers, for example, the contact between the component and the supporting and guiding element, the sound waves also propagate through the component involved in the contact or the supporting and guiding element involved.

The forces involved in the propagation in these bodies can also be transmitted to adjacent objects and in turn to the air, so that structure-borne sound analysis can quickly become very complex. However, in the transfer path analysis, it is preferable to only consider those parts of the structure-borne noise that make a significant contribution to the total noise at the measurement location. The determination of these proportions is preferably part of the analysis. The total noise at the measurement location is preferably made up of airborne and structure-borne noise.

It can also be preferred that the structure-borne noise refers only to the portion of the noise transmitted by solid bodies. As mentioned above, a transfer path analysis can also be performed on the basis of theoretical considerations. In particular, however, the transfer functions of airborne and structure-borne noise should be determined experimentally. In particular, a transfer path of the airborne sound can be analyzed by using a known sound source and at least one microphone or, if the sound source is not precisely characterized, by at least one microphone at the sound source and one at the measurement location. A transfer path analysis of structure-borne noise can be performed by a similar basic principle, but using, for example, accelerometers (acceleration sensors) on the body surface instead of microphones. Accelerometers can preferably be used to determine the vibrations of the body that are characteristic of structure-borne noise

In a preferred embodiment, the method is characterized in that transfer functions of at least three transfer paths are determined. The transfer paths are preferably described using transfer functions, with the at least three transfer paths preferably being defined as airborne noise from a component, structure-borne noise in contact between a component and a supporting and guiding element, and structure-borne noise in contact between a supporting and guiding element and a floor (cf . among other things 2a ). Each transfer path has a sound component of a total sound generated by the calibration system, so that the sum of all transfer paths preferably results in a total sound.

Furthermore, the multiplicity of recorded transfer paths can be viewed in their interaction as a transfer path model. Accordingly, for example, three recorded transfer paths can be viewed together as a transfer path model. The particular advantage of the transfer path model is that with a recorded input signal (excitation signal) for a single transfer path, a signal can be calculated via the transfer function which corresponds to a receiver signal - i.e. the overall sound. $$GesamtscHall=a\mathrm{and}fGenOmmenesVib\mathrm{right}atiOnssiGnal\cdot \ddot{u}be\mathrm{right}t\mathrm{right}aG\mathrm{and}nGsf\mathrm{and}nktiOnen$$

Consequently, by detecting structure-borne noise, e.g. in contact between the support and guide element and a floor - as an excitation signal - on the basis of the transfer path analysis, conclusions can be drawn about the total noise without interference. This can be due in particular to the known transmission functions of the transfer paths happen. The overall sound can be calculated or simulated in this way. In this case, the simulation preferably also includes a simulation of the airborne noise occurring at the measurement location. This is again possible due to the known transfer function for the transfer path of the airborne sound.

In a further preferred embodiment, more than three transfer paths can also be preferred--depending on the calibration system structure.

In a further preferred embodiment, the transfer path model is verified by a microphone measurement. In this case, the first component included in the calibration system can preferably be set in motion, with this being continuously in contact with the first support and guide element. A microphone is preferably placed at a fixed location and records the overall sound, which results in the sum of the sound that is obtained through all transfer paths. At the same time, at least one detection sensor, preferably an acceleration and/or force sensor, records the vibration in the first support and guide element caused by the movement of the first component and selects it as an input variable for the transfer path model. As already described, the overall noise can be simulated using the transfer path model based on the determined transfer paths when recording structure-borne noise in the support and guide element. The simulated overall sound is then compared and verified with the sound recorded by the microphone. If the simulated overall sound differs significantly from the measured sound of the microphone, the transfer paths of the calibration system are preferably re-recorded. The verification can preferably also be carried out at a
stationary first component, in which case the first component is excited by a ram or hammer. Here, too, the measured total sound should preferably be in a similar range to the simulated or calculated total sound.

In another preferred embodiment, a synthesis model is generated by inverting the transfer path model. As already discussed, the transfer path model is preferably defined by one or more transfer paths, each with at least one transfer function H(ω). The sum of the simulated or calculated sound of all transfer paths with their transfer functions represents the total sound emitted. The transfer path model therefore makes it possible to calculate/simulate a total sound when recording an input variable of a single path. The synthesis model preferably reverses these transfer paths—ie the respective transfer functions H(ω)—and accordingly has inverse functions of the transfer functions. Based on a simulated (or recorded) overall sound as an input variable, this makes it possible to determine the individual sound variables of the components involved in the calibration system. The synthesis model therefore preferably also enables a synthesis/extraction of the individual reverse transfer paths. In the context of the invention, the individually determined sound quantities can be viewed as an extracted vibration spectrum.

Based on the synthesis model, the acoustic radiation of the component can be determined.

In a further preferred embodiment, the transfer path analysis can preferably also be described as follows: First, individual transfer paths, preferably on a calibration system, and the associated transfer functions (from the excitation to the receiver) are determined or analyzed. An overall sound is then preferably calculated/simulated at the receiver, with all transfer functions being summed up and with a pulse recorded by a sensor at an excitation point being used as the input variable. In a next step, the simulated overall sound is compared with a microphone measurement so that the transfer paths are validated and, if necessary, the transfer functions of the transfer paths are then adjusted.

A preferred transfer path is extracted via a synthesis model and vice versa, so that an analysis of one transfer path can lead to the identification of the resonance frequency and damping of the crack or damage.

After a transfer path model and a synthesis model have been generated using the calibration system, further components can preferably be checked for damage. Furthermore, within the meaning of the invention, the second component is preferably moved along a longitudinal axis of a second carrying and guiding element, the second component being in continuous contact with the second carrying and guiding element. The second component is preferably the same type of component as the first component, although it is not clear in advance whether the second component is a good part (i.e. a component without damage) or a bad part (a component with damage such as a crack ) acts. The first and the second component have darü In addition, the same geometric dimensions are preferred. The second support and guide element is to be distinguished from the first support and guide element. It is preferably two different support and guide elements. In a further preferred variant, however, the first and second support and guide element can also be the same support and guide element.

In addition, at least one detection device is preferably mounted in the second support and guide element. This at least one detection device preferably picks up vibrations that are generated during the movement of the second component in the second support and guide element. The detected vibrations are then transformed using the transfer path and synthesis model in such a way that an extracted, precise vibration spectrum assigned to the second component is obtained. Ultimately, the detected vibrations of the second support and guide element are selected as the input variable for a damage analysis to be carried out on the second component, and the total sound power is first determined using the transfer path model. Then the synthesis model, which represents the reversal of the transfer path model (i.e. the individual transfer paths), generates or simulates a vibration spectrum from the second component based on the overall sound (cf. also 2a and 2 B ).

In other words, the input variable for the aforementioned signal synthesis is preferably a measurement of the structure-borne noise excitation at the contact point between the second component and the second support and guide element. A sensor (detection device) is preferably used for this, which can measure the force excitation at the point of origin in all spatial directions. Due to the high impulse forces that act at the point at which the excitation occurs, this variable can preferably be determined by measuring other parameters. This includes both the bending of the support and guide element foot (support and guide element - floor element force) and the accelerations on the support and guide element in all three spatial directions and other functions to be integrated (acceleration, pulse measurements, temperature, etc.).

The vibration spectrum of the second component is preferably subjected to an analysis in a final step, with the analysis being carried out with regard to natural frequencies, damping and/or an amplitude, and as a result a crack and/or damage being able to be detected. Based on natural frequencies, damping and amplitude, a special damage pattern can be identified, as this is characteristic of a component in connection with a type of damage.

The analysis is preferably carried out using a data processing unit. The analysis is therefore preferably to be regarded as a computer-implemented method step. Such an analysis can preferably be presented as a comparison between reference data or can also be carried out by an analysis using artificial intelligence algorithms. The analysis of the vibration spectrum is not limited to this.

For example, reference data with known component damage can be generated on the basis of validation tests and the distinguishing features can be extracted using machine learning methods. These then allow the component damage listed to be detected, preferably in the acoustic resonance test. Furthermore, algorithms are preferred with which crosstalk and interference signals can be compensated

• - wobei zwei oder drei Übertragungsfunktionen pro Transferpfad ermittelt werden;
• - wobei die Übertragungsfunktionen eine Reaktion einer Komponente auf eine Kraftanregung und/oder einen Übergang von einer Strukturschwingung in Luft und/oder eine Änderung eines Körperschalls bei einem Übergang zweier Elemente durch Kontaktdämpfung beschreiben.

In a further preferred embodiment, the method is characterized in that the transfer functions are obtained in that the first component is not subject to any movement and the first component and/or the first support and guide element is excited to vibrate via a power impulse hammer and a response signal is measured via an acceleration and/or force sensor and/or a microphone,

• - where two or three transfer functions per transfer path are determined;
• - Wherein the transfer functions describe a reaction of a component to a force excitation and/or a transition from a structural vibration in air and/or a change in structure-borne noise during a transition of two elements through contact damping.

Using this preferred procedure, particularly error-free and resistant transfer functions can be generated, which enable extremely realistic simulations and also do not show any major deviations from the microphone verification at an early stage. An adaptation of the transfer function after a microphone verification is therefore advantageous not to be carried out frequently. In addition, the preferred embodiment also allows transfer functions to be included in a laboratory or in a closed room (preferably also with small dimensions) because the first component is not set in motion. The environmental parameters can be influenced and controlled in a particularly advantageous manner in a closed room or laboratory, which leads to particularly error-free transmission functions. Furthermore, simple means such as a power impulse hammer, in particular ones that are easy to obtain, can be used in order to stimulate the calibration system and, as a result, to generate one or more transfer functions.

In terms of the invention, the first component is preferably not subject to any movement. It goes without saying that the first component is therefore at rest. A body is basically at rest if it does not change its position in relation to a reference body or reference system. In the present case, the position of the first component does not change in relation to the first supporting and guiding element.

A power impulse hammer is preferably selected in order to obtain signal excitation and, as a result, a transfer path analysis on a calibration system. An impulse hammer is preferably a hammer equipped with a force sensor for exciting an object or a first component or a system component in a calibration system with an impulse. The transmission of this impulse in the component can be examined by additional sensors attached to the same component and/or system component. In the present case, impulse hammers are preferably used to excite structure-borne noise. In order to determine the transmission behavior of certain system components (calibration system) - above all in the transfer path analysis - a system component, preferably the first component, is excited with the impact hammer with a force curve that is known (from the force sensor). The vibration of the component or a system component can be determined with structure-borne noise sensors (preferably acceleration sensors), and the emitted airborne noise can be determined with microphones. The transmission behavior of the first component or of the calibration system can be calculated from these measured signals.

Strategic points are preferably excited with so-called power impulse hammers and response signals are measured with acceleration sensors, force sensors and microphones. From this, the transmission functions (FRF) for structure-borne and airborne noise can be determined.

In a further preferred embodiment, a transfer path comprises two or three transfer functions. The transfer function preferably describes the change in structure-borne noise during the transition of two elements through contact damping, the reaction of a component to a force excitation describes the transition from structural vibration to air (without being limited to this). Based on these transfer functions, particularly detailed transfer paths and, associated with this, a transfer path model can be generated.

1. a. das erste Bauteil ein erstes spurgebundenes Rad ist und das zweite Bauteil ein zweites spurgebundenes Rad ist, wobei das erste und zweite spurgebundene Rad einen, bevorzugt den gleichen, Radumfang aufweisen und;
2. b. das erste Trag- und Führungselement eine erste Schiene und das zweite Trag- und Führungselement eine zweite Schiene ist.

In a further preferred embodiment, the method is characterized in that

1. a. the first component is a first fixed wheel and the second component is a second fixed wheel, the first and second fixed wheels having a, preferably the same, wheel circumference and;
2. b. the first support and guide element is a first rail and the second support and guide element is a second rail.

A large number of advantages are achieved by the method according to the invention, in particular with the checking in connection with train components. On the basis of the innovative design of the method, cracks and loosening of a wheel tire can advantageously be detected when a train is passing at cruising speed. The method is to be regarded as a particularly robust (against errors), cost-effective and simple measuring method (and/or system) with little maintenance. In addition, the needs-based maintenance of the wheel tires according to the invention advantageously avoids unnecessary downtimes of a vehicle fleet. In addition, automated and continuous damage detection increases operational safety and reduces the risk of serious accidents caused by loose or broken tires. The preferred method can identify necessary repairs at an early stage, optimize the maintenance intervals of the wheelsets and minimize the effects of damage to the track infrastructure. Every test or workpiece control must clarify whether the test objects are able to withstand the future operational stress without failure. Not only can material properties be determined, but sporadic defects can also be found and evaluated.

The vibration signals of the wheel can be found in the rail through the coupling in the wheel-rail contact. Via a transfer path analysis and synthesis, the vibro-acoustic are preferred physical transmission paths (force excitation, structure-borne noise or structural noise and airborne noise). If the transmission paths and their mathematical composition are known, it is possible in the second step to extract/separate the vibration component of only the wheel, measured with the vibration on the rail near the wheel-rail contact as the input variable. The evaluation of the separated vibration spectrum of the wheel then preferably enables the acoustic signal to be assigned to known damage patterns. An identification of defects is thus advantageously possible.

• 1) Experimentelle Bestimmung der vibro-akustischen Transferpfade (Transferpfadanalyse)
• 2) Erstellung eines parametrischen Transferpfadmodells
• 3) Synthetisierung eines äquivalenten Schallsignals aus Körperschall und Luftschall

The transfer path analysis and synthesis (TPA/TPS) is preferably a test-based method for measuring and calculating the transmission paths of mechanical vibrations. The goal is to analyze the contribution of each individual path from a source to a receiver (e.g. a microphone). The typical procedure for TPA/TPS preferably consists of three steps:

• 1) Experimental determination of the vibro-acoustic transfer paths (transfer path analysis)
• 2) Creation of a parametric transfer path model
• 3) Synthesis of an equivalent sound signal from structure-borne and airborne noise

In a further preferred embodiment, after an initial experimental detection of the transfer paths, a parametric transfer path model is preferably created. This transfer path model enables a synthesis of the relevant sound transmission paths. The overall sound result (sum of airborne and structure-borne noise) can preferably be calculated and verified via a microphone measurement. Furthermore, the validated transfer path model is preferably converted and recalculated in order to extract the vibration contribution of the second wheel alone from the vibration signal at the sensor of the second rail by means of the transfer path synthesis. The vibration signal from the second wheel can then be used for the acoustic resonance test.

Differences between a defective (preferably second wheel) and an intact wheel (preferably first wheel) can preferably be detected by measurement. The wheel oscillates as a common system with the rail and other attachments, such as B. sleepers and clamps. The vibrations of this wheel-rail system are recorded by several vibration sensors attached to the rail.

It can be preferred that the wheel-rail system is particularly important for the transfer path of the structure-borne noise and that its transfer function is the same regardless of the respective train and the location of the rail case. A wheel-rail system preferably includes the wheel involved in the contact (first or second wheel) and the rail involved (first or second rail). Thus, knowledge of a transfer function of the structure-borne noise of a standard wheel-rail system can be sufficient to essentially know the transfer function of all wheel-rail systems.

However, it can also be preferred to individually determine the transfer functions of different wheel-rail systems, preferably at different measurement locations. The transfer function of structure-borne noise is preferably less generalizable for different wheel-rail systems and/or measurement locations than that of airborne noise. The individual determination of the transfer function of the structure-borne noise can then preferably be determined by individual measurements of different wheel-rail systems and/or measurement locations. This is preferably simple for structure-borne noise, since external interference is significantly lower than for airborne noise and preferably no laboratory conditions are required. It may therefore be preferable to install suitable sensors and/or processors for measuring the transfer function at the measurement location and thus essentially measure the transfer function individually.

The measuring method is non-destructive and detects cracks and loosening of the wheel tires of railway wheels when crossing at normal travel speed. Depending on the type of train (night train, freight wagon, tram), this can range from 50 to 140 km/h. The sensors are preferably built into the track (first or second rail) at a planned measurement location and remain there permanently. Once set up, the system works automatically and preferably continuously checks the wheel tire condition of the vehicle fleet.

In a further preferred embodiment, the method is characterized in that the detection device is an acceleration sensor and four acceleration sensors are attached to the second rail at a distance of ¼ wheel circumference along the longitudinal axis of the second rail. The acceleration sensors can preferably be sampled at a rate that is sufficient to provide enough data for a reliable evaluation of the railway wheels passing over them (second wheel) to deliver. The sensor system also enables precise determination of the vibration at the wheel-rail contact. This is the prerequisite for the extraction of the vibration signal purely radiated from the wheel by the applied technology of transfer path analysis and synthesis and the subsequent acoustic resonance test. Acceleration sensors have the advantage of relatively low unit costs (mass production) and high reliability (some such sensors can still withstand accelerations of up to a thousand times the measuring range without damage). Due to their small size, they are also characterized by high measuring speeds.

In an alternative preferred variant, only one acceleration sensor is attached to the second rail. It has been shown that this can be sufficient for a transfer path analysis, with the sensor preferably recording vibrations in only one spatial direction, more preferably in 3 spatial directions. For the calculation of the wheel geometry (polygonization), on the other hand, at least 4 sensors are preferably required.

Wheel noise is mainly caused by out-of-round (polygonized) wheels. With four acceleration sensors included in the second rail, four signals are recorded accordingly. Since the sensors are mounted at a distance of ¼ wheel circumference from one another along the longitudinal axis of the second rail, they interact with one wheel revolution, i.e. one entire wheel circumference of the second wheel. This advantageously generates an image of the polygonization.

The sensors are preferably installed in the track or in the second rail with a defined distance (¼ circumference of the wheel). In addition, the piece of track or the second rail on which the wheel to be measured runs is structurally modified in such a way that it is ensured that the entire wheel tread surface within the measuring section was in contact with the track and that all possible cracks have therefore run through the load area. This is the only way that the system can detect any cracks that may be present or that the wheel tire is not properly seated.

In a further preferred embodiment, the acceleration sensors are attached to the second rail at a distance of preferably ½ wheel circumference from one another along the longitudinal axis of the second rail. Such a distance between the acceleration sensors advantageously reduces the number of acceleration sensors to be used. In fact, only two sensors are required to record a full revolution of the wheel. Although the relatively small number of sensors reduces the accuracy of the measurement results (compared to four sensors, for example), it is to be regarded as sufficient to be able to record a vibration spectrum, which leads to the identification of damage.

In a further preferred embodiment, the acceleration sensors on the second rail are spaced apart at a distance of preferably 1/6, 1/8 or 1/10 of the wheel circumference. With a distance of 1/6 wheel circumference, preferably six sensors are used, with a distance of 1/8 wheel circumference preferably eight acceleration sensors and with a distance of 1/10 wheel circumference preferably ten acceleration sensors second rail a complete wheel revolution can be accommodated. The more sensors are used, the more detailed the measurements and analysis become possible.

The detection devices (acceleration sensors, force sensors, etc.) are preferably to be understood as measuring devices. Strictly speaking, a sensor (also a detector or measuring transducer or measuring sensor) is a technical component and is preferably the first element in a measuring chain in a measuring device. A measuring device or detection device can also preferably include a computing unit and a communication unit as well as a memory unit. In the present document, the terms acceleration sensor, force sensor (or all sensors) mean measuring devices that include the respective sensors.

Also preferably, a wheel detector is attached to the second rail. This wheel detector is preferably a sensor selected from the group consisting of an acceleration sensor, a force sensor, and an optical sensor (without being limited to these). The wheel detector can include any type of sensors that are able to detect an overriding track-bound wheel (preferably a second wheel). The wheel detector is preferably attached to the second rail before (in terms of time and location in relation to a direction of movement of the second wheel) a measurement section, with a measurement section being understood within the meaning of the invention as the acceleration sensors arranged at a distance from one another. The wheel detector can preferably serve as a so-called trigger, which emits a signal after detecting the second wheel, whereby the other acceleration sensors are switched on, whereby the acceleration sensors remain in an energy-saving mode - as long as there is no train or second wheel in the vicinity. In this respect, the wheel detector is preferably placed at a sufficient distance from other components in the environment to enable reactivation of these components in response to an approaching second track-bound wheel. This placement can be made depending on the maximum speed at which the second track-bound wheel traverses the measurement area and a period of time required for the reactivation of the corresponding components.

• - die Beschleunigungssensoren in Datenverbindung mit einer Datenverarbeitungseinheit stehen und die erfasste Schwingung als Rohdaten an die Datenverarbeitungseinheit übertragen wird

und/oder

• - die erfasste Schwingung von Übersprechen und Störsignalen bereinigt wird.

In a further preferred embodiment, the method is characterized in that

• - the acceleration sensors are in data connection with a data processing unit and the detected vibration is transmitted as raw data to the data processing unit

and or

• - the recorded vibration is cleaned of crosstalk and interference signals.

The data processing unit implemented in the method preferably enables the analysis steps of the recorded vibration image to be automated. It is also crucial that the data processing unit has sufficient capacities to enable processing of a large amount of data, the data processing unit being able to carry out an extremely detailed analysis, revealing detailed damage patterns which may not be found by specialist personnel. Due to the fact that raw data are preferably transmitted to the data processing unit without being processed, the detection devices can have a particularly simplified design. Accordingly, the detection devices only require a measuring sensor and a communication unit and can dispense with a memory unit and/or processor.

In terms of the invention, a data processing unit preferably includes means for generating, processing, storing, sending and receiving data. The data processing unit is preferably in data connection with one or more detection device(s) (acceleration sensors, force sensor, microphone, etc.), whereby data can be transmitted bidirectionally between these system components. In a preferred embodiment, the data processing unit carries out algorithms and calculations by receiving input data from the detection devices and generating output data after the algorithms have been carried out, which include a classification of damage. The advantage of such an arrangement is that the detection device does not have to be equipped with components for data processing (processor) and data storage. A further advantage is that the data processing unit can include a large number of recorded data from different acquisition devices for its analysis or implementation of the algorithms, so that a far-reaching analysis is made possible. The data processing unit can preferably be in the form of a server which is connected to the Internet and which other end devices can access. In this way, different responsible bodies or persons can advantageously be informed at any time about the status of a train.

In a further preferred refinement, the data processing unit can control the detection devices. In this case, it can be preferable for all detection devices to remain in a power-saving mode while no train is approaching a measurement section. However, if the wheel detector according to the invention detects a moving wheel, it can transmit a trigger signal to the data processing unit. This can then control the other detection devices in such a way that they switch from a power-saving mode to an operating mode in order to be able to absorb vibrations in the rail.

Raw data is preferably also understood to mean data that has been converted from analog data into digital data via an A/D converter.

Since the second wheel as a component of a train does not stand alone on the second rail, but excites the second rail in combination with other wheels, the measured vibration behavior is preferably corrected for the influence of vibrations of the rail with other attachments. In terms of the invention, this is to be regarded as crosstalk from other wheels and sound paths. Furthermore, other interference signals can also influence the recorded vibrations, which must also be cleaned from the signal.

In a further preferred embodiment, the method is characterized in that the first and/or second rail is structurally adapted such that an entire wheel running surface of the first and/or second track-bound wheel is in contact with the first and/or second rail. In this case, the surface of the first and/or second rail has an arc in cross section, which is designed in such a way that the wheel running surface and the entire rail surface are in contact with one another. Although this generates increased friction between the first and/or second rail and the first and/or second wheel, this can be neglected since the measurement section is not large. By adjusting the splint, an increased level of detail in damage detection can be achieved. This ensures that the entire wheel tread was in contact with the track within the test section and that all possible cracks ran through the load area. This is the only way that the system can detect any cracks that may be present or that the wheel tire is not properly seated.

In a further preferred embodiment, the method is characterized in that the second component is assigned an identification number and/or a warning signal is triggered by the data processing unit in the event of a crack and/or damage. Due to the fact that an identification number is assigned to the second component, the damage can advantageously also be assigned to the identification number and rectified immediately afterwards—after damage has been detected. In this case, it is preferable for the identification number to be assigned in a legally secure manner and only once, so that mixing up components is avoided. Furthermore, an identification number advantageously enables improved archiving of detected damage. The warning signal from the data processing unit can advantageously be used to alert train occupants, train drivers or bodies responsible for repairs, or even emergency services. The warning signal can preferably be an acoustic signal that is emitted to the environment by means of the data processing unit, or else data with a warning are transmitted to terminals that are assigned to the designated competent authorities or persons.

In a further preferred embodiment, a barcode for identifying the wheel is attached to each carriage. The wheel detector is preferably also set up to read this barcode and transmit it to the data processing unit.

Such an arrangement is also generally possible for components, with a workpiece carrier (end of line test) including a barcode for identifying a workpiece component.

In a further preferred embodiment, the method is characterized in that the extracted vibration spectrum is examined for features in an acoustic resonance test and/or the analysis of the vibration spectrum of the second component is carried out using machine learning algorithms.

The data processing unit preferably includes a software module for transfer path synthesis, which transforms the measurement data instantaneously by means of inverse calculation of the parametric transfer path model and interference signal filter. The treated measurement data are preferably cleaned of interference and only reflect the vibration spectrum of the second wheel alone. The acoustic resonance test is preferably carried out on the data transformed in this way in a further software module. This preferably compares the vibration spectra with stored characteristics that indicate a wheel tire crack or a loosening of the press fit. The features can preferably be created using training data. The results of this comparison test are preferably linked to associated vehicle numbers and stored in the operator's maintenance management system. If damage is detected, an alarm is preferably triggered.

In a further preferred embodiment, the method is characterized in that the vibration spectrum of the second component is analyzed using artificial intelligence algorithms. The use of artificial intelligence (in short: AI) to analyze data leads to significant advantages over analysis by conventional (computer-implemented) methods and/or also compared to manual analysis by a human observer. In this way, an AI can advantageously automatically analyze extremely large amounts of data in a very short time. Furthermore, the artificial intelligence algorithms can recognize patterns and/or features in a data set that are not recognized by a human or conventional algorithms. In particular, this means that the AI can detect damage and/or cracks occurring on a second component and/or wheel at an early stage.

In a further preferred embodiment, the artificial intelligence algorithms preferably comprise machine learning algorithms (or machine learning algorithms). In the present case, the term sequences "algorithms of machine learning" and "machine learning algorithms" are used synonymously. It goes without saying that machine learning algorithms are a part of artificial intelligence. Machine learning uses mathematical and statistical models to "learn" from databases. In general, machine learning algorithms have the advantage that information that is too complex for a human observer can be automatically extracted from a large data set. There are a large number of machine learning algorithms that can essentially be divided into three different learning methods: supervised learning, unsupervised learning and reinforcement learning.

According to the invention, methods of supervised learning are particularly preferably used for the analysis of the extracted vibration spectrum. In the supervised learning process, a so-called training process is carried out first. Training data is provided in the form of input data together with the corresponding target data. The purpose of training in machine learning methods is generally to adapt parameters of a function in such a way that the function is then able to determine the target value from the corresponding input value with high accuracy. The fitted function is then used after the training process to predict target data for previously invisible input data. The function is described by a mathematical and/or statistical model.

In a preferred embodiment, the function is configured using a support vector machine, Bayes networks and/or decision trees. The function is particularly preferably described by an artificial neural network. According to the invention, the artificial neural networks can have different architectures and e.g. as Deep Feed Forward (DFF) Network, Recurrent Neural Network (RNN), Deep Convolutional Network (DCN), Deconvolutinal Network (DN), Convolutional Neural Network (CNN), Deep Residual Network (DRN), Boltzmann Machine, Time Delay Neural Networks (TDNNs).

In terms of the invention, the input data are preferably defined by the extracted vibration spectrum.

According to the invention, target data is preferably defined by the classification of the input data in a specific class and/or the occurrence of specific target data (events) is determined on the basis of the input data. Likewise, probabilities for belonging to a specific class can be output as target data or probabilities for the occurrence of specific target data (events). The classes are preferably divided into specific types of damage. For example: "having a crack"; "having a flat spot"; "Fractures and loosening of the interference fit on the wheel disc"; "Have a hairline crack", "No damage". As a so-called classification algorithm, the AI can therefore output whether a specific damage event occurs based on the assignment of the input data to corresponding target data.

In a further preferred embodiment, the unsupervised learning method is used for the analysis or processing of the extracted oscillation spectrum. In unsupervised learning, the algorithm tries to recognize patterns in the input data that deviate from structureless noise. The function in the training process is based only on the similarities of the input data and adjusts its parameters accordingly, so that no output data is used for the training process.

In a preferred embodiment, the unsupervised learning method is used to segment or cluster the input data or, preferably, to compress the input data. The terms clustering, segmentation and compression in connection with machine learning methods are known to those skilled in the art.

In a preferred embodiment, the unsupervised learning algorithm preferably comprises principal component analysis (PCA) and/or the K-Means algorithm and/or at least one neural network.

The proposed method also preferably has artificial intelligence algorithms with unsupervised learning methods. In this context, a dataset of the extracted Vibration spectrum can be pre-processed using unsupervised learning methods, for example by grouping (clustering) relevant data in order to then use this "filtered" data to obtain a classification algorithm for determining a damage event. Such a method can, for example, advantageously filter out the crosstalk or interference signals.

As already described, so-called training processes are carried out in a first step in both of the above-mentioned methods to determine optimal parameters of a machine learning function mentioned above. Based on the fitted function, various statements are made for previously unknown input data after the training.

In a further preferred variant, the method of reinforcement learning is used for the analysis or processing of the recorded process variables. In the case of reinforcement learning, on the other hand, the training process takes place continuously even after the parameters of a function have been adjusted. The effects of different statements are observed and evaluated via "trial and error" using the adapted function for previously unknown input data. In response to these statements, the algorithm receives feedback, presented in the abstract in the form of a reward or punishment. The algorithm then further optimizes the function based on its parameters. Accordingly, the algorithm continuously adapts or changes the function of the machine learning process.

In a preferred embodiment, the method of reinforcement learning includes the Q-learning method and/or the aforementioned neural networks and/or other neural networks and other algorithms known to those skilled in the art

1. a. die mindestens eine Erfassungseinrichtung an einem Trag- und Führungselement befestigt und dazu eingerichtet ist,
   • - eine Schwingung im Trag- und Führungselement aufzunehmen;
   • - die aufgenommene Schwingung als Rohdaten an die Datenverarbeitungseinheit zu übertragen;
2. b. die Datenverarbeitungseinheit dazu eingerichtet ist,
   • - anhand eines bauteilabhängigen Transferpfad- und Synthesemodells die aufgenommene Schwingung zu transformieren, sodass ein extrahiertes Schwingungsspektrum eines mit dem Trag- und Führungselement in Kontakt stehenden und sich bewegenden Bauteils erhalten wird;
   • - das Schwingungsspektrum des Bauteils zu analysieren und daraus resultierend einen Riss und/oder einen Schaden zu erkennen, wobei die Analyse im Hinblick auf Eigenfrequenzen, Dämpfung und/oder Amplituden erfolgt;
   • - das extrahierte Schwingungsspektrum in einem akustischen Resonanztest auf Merkmale hin zu untersuchen und/oder das extrahierte Schwingungsspektrum über Algorithmen des maschinellen Lernens zu analysieren.

In a further preferred embodiment, the invention relates to a system for non-contact, non-destructive real-time component monitoring, characterized in that the system comprises at least one detection device and a data processing unit, wherein

1. a. the at least one detection device is attached to a support and guide element and is set up for this purpose,
   • - absorb a vibration in the support and guide element;
   • - to transmit the recorded vibration as raw data to the data processing unit;
2. b. the data processing unit is set up to
   • - to transform the recorded vibration using a component-dependent transfer path and synthesis model, so that an extracted vibration spectrum of a moving component that is in contact with the support and guide element is obtained;
   • - to analyze the vibration spectrum of the component and, as a result, to detect a crack and/or damage, the analysis being carried out with regard to natural frequencies, damping and/or amplitudes;
   • - examine the extracted vibration spectrum for features in an acoustic resonance test and/or analyze the extracted vibration spectrum using machine learning algorithms.

Such a system for solving the above-mentioned problem is neither known from the prior art nor suggested to an average person skilled in the art. Rather, the system according to the invention is to be regarded as a departure from the prior art, in which manual monitoring, analysis, transmission and/or processing steps often have to be included, particularly for damage detection. In contrast, the system according to the invention offers the possibility of carrying out a fully automatically executed method for non-contact, non-destructive real-time component monitoring. Above all, the system also offers the possibility of providing a particularly compact measurement setup, which requires only a few system components and - once it has been calibrated - can be used for a large number of measurements and analyzes without further changes.

A person skilled in the art recognizes that the advantages, technical effects and preferred embodiments discussed in connection with the method according to the invention apply analogously to the system according to the invention for non-contact, non-destructive real-time component monitoring. Likewise, all advantages, technical effects and preferred embodiments that are described in the context of the system can be transferred to the method.

• - Übertragungsfunktionen von drei Transferpfaden ermittelt werden;
• - ausgehend von den drei Transferpfaden ein Transferpfadmodell erzeugt wird;
• - das Transferpfadmodell durch eine Mikrofon-Messung verifiziert wird;
• - das Synthesemodell erhalten wird, indem das Transferpfadmodell umgekehrt wird.

In a further preferred embodiment, the system is characterized in that the component-dependent transfer path and synthesis model is generated by carrying out a transfer path analysis on a calibration system, the calibration system having a component that is free of cracks and/or damage and a component that is in contact with the component supporting and guiding element,
in which

• - Transfer functions of three transfer paths are determined;
• - A transfer path model is generated based on the three transfer paths;
• - the transfer path model is verified by a microphone measurement;
• - the synthesis model is obtained by inverting the transfer path model.

Before installing the system according to the invention, the measuring section is preferably calibrated. For this purpose, a parametric transfer path model is preferably created using the transfer path measurements. The software for this is preferably implemented as a module within the preferred data processing unit. The transfer path model contains the relevant transfer paths and all relevant secondary parameters. For the calibration measurements, one component is preferably placed one after the other on each position of a detection device.

In a further preferred embodiment, the system is characterized in that the component is a track-bound wheel, the track-bound wheel having a wheel circumference and the support and guide element being a rail.

The system is preferably first calibrated and measured for one rail. For this purpose, a transfer path analysis is preferably carried out with a track-bound wheel on the rail in order to determine exactly all transfer paths via the wheel, rail and a sleeper. The results are preferably verified with microphone measurements when the track-bound wheel drives past. Sections in the track network of a railway operator, over which all vehicles in a vehicle fleet to be examined must drive, e.g.

The terms "track" and "rail" are used interchangeably throughout the document.

The created transfer path model is preferably reversed with a subsequent transfer path synthesis in order to draw conclusions about the vibration contributions of the individual components from the vibration signal of only the rail, close to the wheel-rail contact. If this succeeds, the transfer path, which represents the sole contribution of the wheel to the vibration signal, is isolated. The synthesis of this isolated transfer path is preferably implemented in analysis software for the data processing unit.

The vibrations in the rail are preferably measured during the crossing and the measurement data are transformed instantaneously by the analysis software with the implemented synthesis model. Furthermore, filters against crosstalk and interference signals are preferably applied here. The result is preferably the exact, extracted vibration spectrum of just the wheel/wheel tire, without interference components.

The extracted vibration spectrum of the wheel/wheel tire is then examined, preferably in an acoustic resonance test (ART), for features that indicate crack damage or loosening of the wheel tire. The ART is preferably also implemented in software included in the data processing unit and works automatically and instantaneously for the crossing. The distinguishing features for the damage detection are preferably defined using training data. Machine learning methods are also preferably used here. The system preferably has a modular structure and can therefore be adapted to local requirements.

In a further preferred embodiment, the system is characterized in that the detection device is an acceleration sensor and four acceleration sensors are attached to the second rail at a distance of ¼ wheel circumference along the longitudinal axis of the second rail.

The measuring system preferably comprises a number of vibration sensors permanently attached to the rail (wayside system). The sensors are preferably read consecutively by the data processing unit as the wheel passes over them, in order to detect vibrations from the entire circumference of the wheel. The measuring section is also preferably adapted in such a way that all possible contact points of the wheel tread are recorded for each wheel.

CHARACTERS

The invention is to be explained in more detail below with reference to figures, without being restricted to these.

character list

• 1 Schematic representation of a mechanism for generating rolling noise in a preferred embodiment, namely on a wheel and a rail
• 2 Schematic representation of a preferred embodiment, namely a wheel-rail contact and the associated transfer paths of sound
• 3 Schematic representation of a preferred embodiment, namely a wheel-rail contact, particularly showing a preferred system arrangement
• 4 Representation of a transfer path
• 5 Schematic representation of a calibration system preferably having two transfer paths
• 6 Schematic representation of transfer paths of a preferred wheel-rail contact for detecting damage to a wheel
• 7 Preferred transfer path, which is described by three transfer functions
• 8th Schematic representation of transfer paths of a preferred wheel-rail contact for detecting damage to a wheel, taking into account the individual transfer functions

Detailed description of the illustrations

1 FIG. 12 shows a schematic representation of a mechanism of rolling noise generation in a preferred embodiment, namely on a wheel and a rail. In terms of the invention, the wheel is preferably to be understood as a component 1 and the rail is preferably to be understood as a supporting and guiding element 3 .

Vibrations in trains are primarily caused by wheel-rail contact when the wheel rolls on the track. The roughness of wheel and rail preferentially induces vibrations in both wheel and rail. There is a corresponding force excitation at the wheel-rail contact. The method and system according to the invention is based precisely on this mechanism: the vibration signals of the wheel can also be found in the rail due to the coupling in the wheel-rail contact. The vibro-acoustic transmission paths (force excitation, structure-borne noise or structural noise and airborne noise) can preferably be found via a transfer path analysis and synthesis. If the transmission paths (transfer paths) and their mathematical composition (transfer functions) are known, it is possible to extract/separate the vibration component of the wheel, with vibration on the track near the wheel-rail contact being measured as the input variable. Finally, the evaluation of a separated vibration spectrum of the wheel enables the acoustic signal to be assigned to known damage patterns, which makes it possible to identify defects.

2 ( 2a and 2 B .) illustrates a preferred embodiment, namely a wheel-rail contact, and the technical procedure of a transfer path analysis and synthesis. In terms of the invention, the wheel is preferably to be understood as a component 1 and the rail as a supporting and guiding element 3 . Furthermore, a detection device 5 is preferably attached to the rail, which is set up to record vibrations in the rail.

the 2a shows in particular schematically a transfer path analysis for the preferred wheel-rail system. In 2a it is preferably a calibration system with a first wheel and a first rail. First, the transfer paths of the structure-borne and airborne noise transmission are measured in the calibration system, which result in the total noise without interference. In this case, an analysis of all transmission paths (transfer paths) is preferably carried out on a stationary (that is to say not showing any movement) first wheel, with this wheel showing no damage or cracks. The aim of this analysis is to obtain information about the impact of force impulses in all three spatial directions on the structure-borne and airborne noise spectrum at a target location. For this purpose, strategic points are excited with so-called power impulse hammers and response signals with detection devices 5 (preferred Acceleration sensors, force sensors and/or microphones) measured. From this, transfer functions (FRF) for structure-borne and airborne noise can be determined, each of which is assigned to a transfer path. A transfer path can preferably include two or three transfer functions. A first transfer path preferably describes an airborne noise from the first wheel, a second transfer path describes structure-borne noise in the wheel-rail contact and a third transfer path describes structure-borne noise in the sleeper track bed contact. In abstract form, the track bed can preferably also be understood as a floor.

After the initial experimental recording of the transfer paths (transfer path analysis), a parametric transfer path model is created. With this model, the total sound can be obtained with a real excitation (e.g. by a rolling train with a second wheel on a second rail), whereby the excitation of a single transfer path is sufficient as an input variable for the transfer path model.

In 2 B a real train crossing is shown schematically, with a second wheel and a second rail being illustrated. As the train passes, the second wheel moves and is in constant contact with the second rail. Due to the movement, vibrations are induced in the wheel and rail alike. Through the above (p. 2a ) Transfer path model, a synthesis model is preferably generated, it being possible to back-calculate the pure airborne noise contribution of the individual second wheel using the synthesis model, starting from an overall noise determined by the transfer path model. The required input variable for this signal synthesis is a measurement of the structure-borne noise excitation at the point of contact between rail and wheel. For this purpose, a preferred sensor (detection device 5) measures a force excitation at the point of origin in all spatial directions. The pure airborne noise contribution of the individual second wheel is to be understood as an extracted vibration spectrum, which can preferably be subjected to further analyzes in order to identify damage and/or cracks on the second wheel.

3 FIG. 12 shows a schematic representation of a preferred embodiment, namely a wheel-rail contact, particularly showing a preferred system arrangement. The preferred system has four detection devices 5 and a data processing unit 7 . The system is preferably set up to carry out non-contact, non-destructive real-time component monitoring. In particular, the system is set up to analyze a train crossing a track in order to identify a crack and/or damage in a track-bound wheel as a result.

The train has a second component 1, namely a second wheel, which is in contact with a second supporting and guiding element 3, namely a second rail, the second wheel being in motion on the rail and the contact being maintained throughout is obtained. The detection devices 5 of the preferred system are installed in the track (second rail) at a defined distance from one another, namely preferably ¼ the circumferential length of the second wheel. In addition, the second rail, on which the second wheel to be measured runs, is structurally modified in such a way that it describes a lateral arc. This ensures that the entire wheel tread is in contact with the rail within the test section and that all possible cracks have run through the load area.

Since the second wheel as part 1 of a train does not stand alone on the second rail, but excites the second rail in combination with other wheels, the measured vibration behavior still has to be adjusted for the influence of vibrations of the rail with other attachments (crosstalk from other wheels and sound paths). For this, a detailed investigation of the transfer paths of the wheel-rail system must be carried out at the measuring point (or a calibration system).

Furthermore, the system can be expanded to include a wheel detector 11 mounted in the second rail, with the wheel detector 11 preferably comprising an acceleration sensor and increasing the accuracy of the measurement results. The vibration signals (oscillations) determined by the detection devices 5 in the second rail are transmitted to the data processing unit 7, these being converted into a digital signal beforehand, preferably via an A/D converter 9. The data processing unit 7 is preferably set up to transform the recorded vibration using a wheel-dependent transfer path and synthesis model, so that an extracted vibration spectrum of the second wheel is obtained. Furthermore, the data processing unit 7 can analyze the vibration spectrum of the second wheel and, as a result, detect a crack and/or damage, with the analysis being carried out with regard to natural frequencies, damping and/or amplitudes. Furthermore, the data processing unit 7 is set up for the extracted oscillation spectrum in an acoustic resonance test for features and/or to analyze the extracted vibration spectrum using machine learning algorithms.

Furthermore, as is known from the prior art, the second rail is fastened to a plurality of sleepers 13 via a clamp 15, as a result of which the loads on the second rail are transferred to a floor and distributed evenly.

4 shows the elements of a transfer path. In order to be able to perceive vibrations acoustically, they have to propagate along a certain path from the point of origin to the point of reception, in this case to an ear of a receiver. These propagation paths are called transfer paths and can be made "visible" using the transfer path analysis. i.e. The task of a transfer path analysis is to identify the flow of energy in the form of structure-borne or airborne noise from one or more sources via different transfer paths to a receiver.

In other words, a transfer path describes the transmission of a signal. If a system is excited to vibrate, for example by a power impulse hammer, the system generates a response signal that reaches a receiver. The system thus transforms an excitation signal into a receiver signal. Such a transformation can be described by transfer functions H(ω), where one or more transfer functions belong to a transfer path. The system therefore corresponds to a division of receiver signal and excitation signal or a division of output signal and input signal: $$\begin{array}{ccc}\underset{\_}{H}\left(\omega \right)=\frac{A\left(\omega \right)}{\underset{\_}{E}\left(\omega \right)}& System=\frac{A\mathrm{and}sGanG}{EinGanG}& System=\frac{Empf\ddot{a}nGe\mathrm{right}}{An\mathrm{right}eG\mathrm{and}nG}\end{array}$$

5 shows a preferred calibration system with two transfer paths. Such a calibration system preferably comprises a number of components, these components each transmitting structure-borne noise to one another. This means that in a transfer path analysis for a system with several components, transfer functions between the components must also be taken into account. In the transition between the components, the sound behaves like structural sound.

Here the preferred calibration system comprises two system components which are connected to one another, i.e. they are in contact with one another. If a first component is excited by a power impulse hammer, the calibration system preferably generates two transfer paths, each of which has a portion of a total sound that reaches the receiver.

6 12 illustrates transfer paths of a preferred wheel-rail contact for detecting damage to a wheel. The wheel-rail contact depicted is preferably a second wheel in contact with a second rail, with the second wheel in motion along a longitudinal axis of the second rail is moved. This causes the rail to experience an unknown excitation. The second rail is also preferably in contact with a threshold that introduces emerging forces into a floor.

Based on a transfer path analysis on a calibration system, which generates a transfer path model, the transfer paths can also be adapted for the existing wheel-rail contact. A single input signal - e.g. an unknown excitation of the second rail by the second wheel - is sufficient to calculate or simulate the total sound to be reached by the receiver using the known transfer functions in the transfer path model. $$Empf\ddot{a}nGe\mathrm{right}=An\mathrm{right}eG\mathrm{and}nG\cdot \ddot{u}be\mathrm{right}t\mathrm{right}aG\mathrm{and}nGsf\mathrm{and}nktiOn$$

7 shows a preferred transfer path, which is defined by three transfer functions. The transfer function "MTL" (Mount Transmissibility) describes the change in structure-borne noise when two elements pass through due to contact damping. The transfer function "AP" (Apparent Mass) describes the reaction of a component to a force excitation. Whereas the transfer function “ATF” (Vibro-Acoustic Transfer Function) includes the transition from structural vibration to air.

8th schematically illustrates three transfer paths of a preferred wheel-rail contact for detecting damage to a wheel, with the individual transfer functions being taken into account in the illustration. In this case, an excitation signal is recorded in one rail (preferably a second rail). taken, which is generated by a moving and standing in contact with the (second) rail second wheel. The detection device 5 for recording the vibration in the (second) rail is preferably an acceleration sensor. This is preferably set up to absorb vibrations in three spatial directions (x, y, z). The vibrations in the second rail recorded by the acceleration sensor make it possible to determine the sound component of the second wheel, the second rail and the sleeper (from the total sound) on the basis of individual transfer paths and their transfer functions. The transfer paths have been determined in a previous step using a calibration system comprising a first stationary wheel and a first rail. The transfer paths, starting from the wheel or from the sleeper to a receiver, each have three transfer functions (MTL, AP and ATF) for each recorded spatial direction (x, y, z) of the vibration. This means that both transfer paths include a total of nine transfer functions. In contrast, the transfer path, starting from the second rail to the receiver, has two transfer functions per spatial direction recorded. Accordingly, the transfer path includes a total of six transfer functions. The respective transfer functions of the transfer paths can also preferably be multiplied by a factor.

Reference List

1

component

3

supporting and guiding element

5

detection device

7

data processing unit

9

A/D converter

11

Wheel Detector

13

threshold

15

clamp

QUOTES INCLUDED IN DESCRIPTION

This list of the documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• US20160207552A1[0010]
• DE 102018123445 A1 [0011]

Claims (10)

Process Applications for non-contact, non-destructive real-time component monitoring, characterized in that the process comprises the following steps: a. Provision of a calibration system comprising a first component (1) and a first support and guide element (3), the first component (1) being free of a crack and/or damage and being in contact with the first support and guide element (3) standing; b. Carrying out a transfer path analysis on the calibration system, wherein - transfer functions of the transfer paths involved are determined; - A transfer path model is generated based on the transfer paths involved; - the transfer path model is verified by a microphone measurement; - a synthesis model is generated by inverting the transfer path model; c. Moving a second component (1) along a longitudinal axis of a second support and guide element (3), the second component (1) being in continuous contact with the second support and guide element (3); i.e. Detection of vibrations in the second support and guide element (3) by at least one detection device (5) attached to the second support and guide element (3); e. Transforming the detected vibrations using the transfer path and synthesis model, so that an extracted vibration spectrum of the second component (1) is obtained; f. Analysis of the vibration spectrum of the second component (1), the analysis being carried out with regard to natural frequencies, damping and/or an amplitude and as a result a crack and/or damage being able to be detected. procedure after claim 1 characterized in that the transfer functions are obtained in that the first component (1) is not subject to any movement and the first component (1) and/or the first supporting and guiding element (3) is excited to vibrate via a power impulse hammer and a Response signal is measured via an acceleration and/or force sensor and/or a microphone, two or three transfer functions per transfer path being determined; - Wherein the transfer functions describe a reaction of a component to a force excitation and/or a transition from a structural vibration in air and/or a change in structure-borne noise during a transition of two elements through contact damping. Method according to one or both of the preceding claims , characterized in that a. the first component (1) is a first track-bound wheel and the second component (1) is a second track-bound wheel, the first and second track-bound wheels having one, preferably the same, wheel circumference and; b. the first carrying and guiding element (3) is a first rail and the second carrying and guiding element (3) is a second rail. procedure after claim 3 characterized in that the detection device (5) is an acceleration sensor and four acceleration sensors are attached to the second rail at a distance of ¼ wheel circumference along the longitudinal axis of the second rail. procedure after claim 4 characterized in that - the acceleration sensors are in a data connection with a data processing unit (7) and the detected vibration is transmitted to the data processing unit (7) as raw data and/or - the detected vibration is cleaned of crosstalk and interference signals. procedure after claim 4 and/or 5 characterized in that the first and/or second rail is structurally adapted in such a way that an entire wheel running surface of the first and/or second track-bound wheel is in contact with the first and/or second rail. Method according to one or more of the preceding claims , characterized in that the second component (1) is assigned an identification number and/or a warning signal is triggered by the data processing unit (7) in the event of a crack and/or damage. Method according to one or more of the preceding claims , characterized in that the extracted vibration spectrum is tested for features in an acoustic resonance test is sought and/or the analysis of the vibration spectrum of the second component (1) is carried out using machine learning algorithms. System for non-contact, non-destructive real-time component monitoring, characterized in that the system comprises at least one detection device (5) and a data processing unit (7), wherein a. the at least one detection device (5) is attached to a support and guide element (3) and is set up to - record a vibration in the support and guide element (1); - To transmit the vibration recorded as raw data to the data processing unit (7); b. the data processing unit (7) is set up to - transform the recorded vibration using a component-dependent transfer path and synthesis model, so that an extracted vibration spectrum of a moving component (1) that is in contact with the support and guide element (1) is obtained; - To analyze the vibration spectrum of the component (1) and, as a result, to detect a crack and/or damage, the analysis being carried out with regard to natural frequencies, damping and/or amplitudes; - examine the extracted vibration spectrum for features in an acoustic resonance test and/or analyze the extracted vibration spectrum using machine learning algorithms. system after claim 9 characterized in that the component-dependent transfer path and synthesis model is generated by a transfer path analysis being carried out on a calibration system, the calibration system having no crack and/or damage in the component (1) and a component (1) in contact with the supporting and guide element (3), wherein - transfer functions of three transfer paths are determined; - A transfer path model is generated based on the three transfer paths; - the transfer path model is verified by a microphone measurement; - the synthesis model is obtained by inverting the transfer path model; wherein the component (1) is a track-bound wheel, wherein the track-bound wheel has a wheel circumference and the supporting and guiding element (3) is a rail; wherein the detection device (5) is an acceleration sensor and four acceleration sensors are attached to the second rail at a distance of ¼ wheel circumference along the longitudinal axis of the second rail.

Images