DE10028519A1 - Quality inspection method for cast or sintered components, involves dividing response frequency spectra into regions of fundamental resonance and above fundamental resonance depending on time - Google Patents

Abstract

The cast component (1) is excited with a broadband acoustic pulse and the response frequency spectra is divided into regions of fundamental resonance and above fundamental resonance band depending on time. The divided spectra is compared with reference using fuzzy classification or neural network to judge the quality of the component.

Description

Technical application area

The present invention relates to a method for quality inspection of components in which the component excited with a broadband acoustic impulse and the impulse response of the component is recorded and a Fre is subjected to sequence analysis in order to determine intact components distinguish defective components.

The process takes place particularly in the area of Quality assurance in production varies ster components, for example from cast or sintered materials, application.

State of the art

It has long been known the quality of construction recognize parts through their sound. It will Component usually struck mechanically and the resulting component resonances via structure-borne noise or airborne sound measurement determined. The comparison typi frequencies with calculated or measured Values then provides information about defects in the component, which are reflected in the change in sound.

Such a method is, for example, from US 5,285,687 for the investigation of monolithic construction parts known. In this process, the component by mechanical striking with a broadband acoustic impulse excited and the impulse response over Airborne sound measurements recorded. The captured data will  a frequency analyzer, with which the distribution sound pressure to individual frequencies within a predetermined frequency range is determined. The result is compared with a reference curve for Damaged components compared, whereby intact damaged components can be distinguished. The Fre quenz range for the examination of the component lies in the example of this document in the audible range between 4 and 12 kHz.

In numerous applications, one of the like method for quality inspection of components each yes or provides unreliable statements. The reasons for this problem can be that the component defects to be determined are not considered fundamental resonances of the component reflect that the excitation and measurement are not made reproducible or that disturbances from the environment in the considered frequency range a reliable statement thwart.

US 5,425,272 describes another method for the quality inspection of components, in which at least two outstanding frequencies of the component over one acoustic excitation can be determined and the ratio the frequency difference to one of the two frequencies zen with pre-calculated or pre-determined values for undamaged components are compared.

Since this as well as other comparable procedures defective components based on a frequency shift of Recognizing resonance frequencies is for an exact measurement solution requires a very high frequency resolution to also very small shifts in these resonances to be able to recognize. However, this leads to a corresponding  long measuring time, which is precisely in the area of Pro production technology is undesirable.

Based on this state of the art The object of the present invention is to provide a method to provide for the quality inspection of components, the a reliable quality check with low demand requirements for frequency resolution and thus with ge allows shorter measuring times.

Presentation of the invention

The task is accomplished with the method according to Patentan spell 1 solved. Advantageous embodiments of the Ver driving are the subject of the subclaims.

In the method according to the invention, this becomes too un component searching, as well as in known methods for the quality inspection of components, with a broad band the acoustic impulse and the impulse response of the component via appropriate sensors. The The present method is characterized in that from the impulse response the distribution of the excitation generators gie on frequency ranges of a broad frequency band the impulse response depending on the time is communicated. The broad frequency band includes ins especially an area above the basic resonances of the component and can also include the basic resonances conclude. A comparison of one or more ren by predetermined frequency and time segments agreed sections from the determined distribution with intact data, defective components can be Components can be distinguished.  

In contrast to the known methods of Stan the technology in which the frequency analysis is time-integrated grail is carried out in the process of before lying invention the distribution of the excitation energy to different frequency ranges or frequencies time-resolved within a broad frequency band examined. This time resolution is preferably about the known techniques of "Joined Time Frequency Ana lysis "- applied to the data of the captured Im pulse response - reached. It goes without saying that due to physical reasons there is a middle ground between time and frequency resolution can be selected in order to ensure that the si to be able to provide significant clippings.

It has been shown that precisely through Untersu of the sound behavior above the predictable and predictable basic resonances of the component in the connection with the time-resolved view very reliable statements about the existence of construction partial defects can be hit. Here are much lower frequency requirements Solution presented as this in the known methods the state of the art is required. The correspond This reduces the measuring time on the components ed.

The specification of the frequency and time segments the recorded frequency and time-resolved distribution the excitation energy depends on the components or test objects to be examined in each case. Here the sections are currently selected, on the most reliable for those defects of intact components can be distinguished. It can be  act a single cutout that also covers the whole recorded frequency-time window can include, or around several different clippings that are given if characteristic for different component defects are stish. A comparison is made during the evaluation with reference values, these reference values preferred have been determined on intact components. It is however also possible, depending on the application, on a Agreement in the respective sections to check lying data pattern with reference data, those recorded on components with certain defects were.

The broad frequency band in which the evaluation he follows, preferably covers a range of min at least 200 kHz. The frequency range can also vary from Range of audible frequencies up to 400 kHz or extend into the MHz range.

The one or more transducers for generating the Excitation pulse and to receive the impulse response are preferably directly mechanical with the component brought into contact. This creates a connection between the component and the sensors to an electro mechanically coupled system achieved. Sensors or The converter and component are used in the quality inspection considered an acoustic unit.

The interpretation of the sensor or converter properties is preferably carried out in such a way that the acoustic Signature of the test object dominates, that is, that the influences of the transducers or sensors on the Im pulse response of the component in the frequency ranges, in  the defects of the component based on certain patterns are not dominant in this pattern.

Furthermore, electrically controllable when used rer converter for generating the excitation pulse Transducers driven by an electrical pulse the one that is the inverse transfer function of this wall lers. This way the transmission Compensate the function of the converter when the excitation Siert.

By coupling the transducers directly to the Component it is necessary to reproduce this coupling designable. This is preferably because of this achieved that the converter (s) and the component oh ne mechanical influence on the component with a defined force are pressed against each other is generated by suitable physical effects. Such physical effects, in which not mecha The component is acted on niches, for example have the gravitational force with which the component on the sensors, adhesive forces between the Sensors and the component, for example by ei a suitable thin liquid film between Senso ren and component can be generated, or magnetic Forces in particular for attaching with magnets equipped sensors or transducers to a metalli cal component can be used.

For the use of the method according to the invention the use of only one converter is also suitable Sender and receiver. With this design lets in particular the quality inspection of very small Execute components with high reliability. Preferably  becomes a piezoelectric transducer used.

Furthermore, a separate converter can also be used Generation of the excitation pulse and one or more use additional transducers to record the impulse response be det. In addition to piezoelectric transducers for detection solution of the impulse response can also be very advantageous use a laser scanning vibrometer.

The comparison of the recorded data with the Refe reference data to differentiate intact components from de defective components are made using a suitable comparison technique techniques, for example via correlation coefficients. Preferably a fuzzy classifier or a new one ronal network, such as a Kohonen network set. Of course, there are also other classifications ornamentation techniques, such as cluster analysis, suitable for use in the present method. For the time-dependent display of the Impulsant short-term FFT Transformation or the wavelet transtormation.

The invention is based on an off example in connection with the drawings again briefly explained. Here show:

Fig. 1 is a schematic representation of the measuring principle for the implementation of the present procedural;

Fig. 2 is a false color representation (here: gradations in Grauab) of a toothed belt pulley vibrating at 1900 Hz;

FIG. 3 shows two Sonogrammdarstellungen of the detected before lying invention acoustic Si gnatur according to two components;

Figure 4 shows the block diagram of the metrological implementation 5 of a signature analysis system according to the vorlie invention.

Fig. 5 is an example of a selection of the program Sono cutouts in the present method; and

Fig. 6 by way of example a schematic of the automated evaluation of Sonogrammausschnitte using fuzzy classification.

Ways of Carrying Out the Invention

Fig. 1 shows two possible arrangements of sensor or transducer 2 , 3 and component or device under test 1 , as they can be used in the implementation of the present method.

In the upper arrangement, component 1 is only brought into contact with a single sensor 2 , which works simultaneously as a transmitter and as a receiver. First, a defined acoustic pulse 4 is introduced via the sensor 2 into the component 1 and then the resulting impulse response of the component, hereinafter referred to as acoustic signature 5 , is received again via the same sensor 2 and an evaluation or processing (not shown) unit supplied.

The second arrangement in the lower area of the figure shows an embodiment in which a sensor 2 as sensor and another sensor 3 as receiver are brought into contact with component 1 . The application of the component 1 with the acoustic impulse 4 takes place via the sensor 2 , the detection of the impulse response via the sensor 3 . The resulting acoustic signature 5 is in turn fed to an evaluation device, not shown, in which the further method steps are carried out.

Fig. 2 shows an example of a snapshot of the vibration speed (rapid) of a vibrating at 1900 Hz measured with a laser scanning vibrometer in false color representation (here: different gray gradation). This is one of the fundamental vibration modes of the sintered metal pulley. Quality deficiencies on the toothed belt pulley are only partially reflected in such basic vibration modes as are examined in the prior art.

To test the quality of such a component using the method of the present invention, an arrangement was created for measuring the broadband impulse response in which the toothed belt pulleys to be tested are positioned in an identical position on the tips of three piezoelectric sensors or transducers using a suitable handling device be reproducibly set up. The tips of the sensors or transducers are arranged in a horizontal plane, so that the toothed belt pulley rests on the sensors with a defined force that is the same from disk to disk due to the gravitational force. One of these sensors serves as a transmitter converter for coupling the excitation pulse, the two remaining sensors are used to record the impulse response. When carrying out the method, a defined high-voltage pulse is given to the transmitter and the broadband pulse response (1 kHz-250 kHz) of the complex arrangement of sensors and component is measured at the other two sensors. The frequency spectra are then calculated from the recorded time signal in short time segments and displayed in colored sonograms. The color shows the current energy in the respective frequency or in the respective (narrow) frequency range. Examples of such sonograms can be seen in FIG. 3.

Fig. 3 shows in the upper area the sonogram recorded and represented by this technique of an intact reference component, in the lower part the corresponding sonogram of a damaged component. These exemplary sonograms have been recorded for a selected frequency range up to 40 kHz and over an observation time of 200 ms. It can be clearly seen from the two representations that the sonograms or acoustic signatures of the two measured components differ significantly. From the representations it can also be seen that the component defect (s) only have a very small effect on the basic resonance of the component of 1900 Hz shown in FIG. 2.

In the sonograms shown, the Recognize color only as a gradation of gray. This is however only for a visual check of this nograms of importance. In automated tests, who not the visual representations, but those to them  the underlying data or data pattern chen.

Provided that to grasp it send component differences in the acoustic signature - in the entire area or only in selected suns The system is very capable of reproducing nogram excerpts advantageous for automated testing of the respective Components are used.

The success of the system is determined by an appropriate one highly reproducible and on the component's own customized individual sensor / component arrangement essentially determined. Here are not for every application uses the same sensors. The Rather, sensors are tailored to the specific requirement individually adjusted. In the present case the sub There were three searches for timing belt pulleys pointed sensors selected.

Fig. 4 shows the block diagram of an example of a metrological implementation of the present signature analysis system. Sonograms are calculated from the impulse response of the system in two frequency ranges (0 -50 kHz and 50-200 kHz). Of course, a division into more than two frequency ranges is also possible. The analog signal detected by the sensors is A / D converted and divided into the frequency ranges above and below 50 kHz using a digital filter with a cutoff frequency of 50 kHz. The respective areas are subjected to Fourier transformations in order to calculate the respective sonogram. In the present example, four sonogram sections are selected, classified in a fuzzy classifier with a reference to be learned, and processed for a good-bad decision based on an individual weighting.

Examples of such selected Sonogrammaus sections are shown in Fig. 5. The figure shows in the upper left area the fully recorded sonogram and in the lower area the four selected time-frequency sections 6 , by means of which good components can be distinguished from bad components in a particularly reliable manner. The appropriate selection of these time-frequency sections was determined by preliminary tests.

Fig. 6 shows an example of a diagram of the program cutouts in the present embodiment durchgeführ th automated evaluation of the selected Sono by fuzzy classification. By comparing the selected sonogram sections of the test objects with the reference sonogram sections, the 16 correlation coefficients are determined, which are added to the fuzzy classification and, after weighting and linking the individual decision, lead to the final distinction between good and bad components.

The evaluation of the recorded impulse response was in the present example with a compact PCI Computer with a signal processing, which on a DSP board with associated data acquisition system reali was carried out. Piezo electrical structure-borne noise sensors of various types, a piezo actuator is used as the transmitter.

Claims (18)

1. A method for quality inspection of components, in which the component is excited with a broadband acoustic pulse and the impulse response of the component is recorded, the impulse response being subjected to a frequency analysis in order to distinguish intact components from defective components, characterized in that the Distribution of the excitation energy to frequency ranges of a broad frequency band which is determined in the pulse response as a function of time, the broad frequency band also or exclusively comprising frequencies above the fundamental resonances of the component, and a comparison of one or more by predeterminable frequency and time segments certain sections of the determined distribution is carried out with reference data in order to distinguish the intact components from the defective components. 2. The method according to claim 1, characterized, that one or more transducers to generate the Excitation pulse and for receiving the pulse answer directly mechanically with the component in con be brought to the beat. 3. The method according to claim 1 or 2, characterized, that the time-dependent distribution by respective  Calculation of the frequency spectra in short successive periods of the pulse answer is determined. 4. The method according to claim 3, characterized, that the time-dependent distribution over a wave let transformation is determined. 5. The method according to any one of claims 1 to 4, characterized, that the broad frequency band has a range of too includes at least 200 kHz. 6. The method according to any one of claims 1 to 5, characterized, that the predeterminable frequency and time segments be chosen in such a way that in the be agreed sections of the distribution significant Changes in the distribution of defective components compared to the reference data. 7. The method according to claim 6, characterized, that the frequency and time segments by pre attempts to be determined. 8. The method according to any one of claims 1 to 7, characterized, that the reference data in the form of a reference sonograms are provided.   9. The method according to any one of claims 1 to 8, characterized, that the comparison of the one or more off cuts from the determined distribution with Refe reference data takes place via a correlation, the Differentiation of the intact components from the de perfect components based on correlation coefficient is carried out. 10. The method according to claim 9, characterized, that a fuzzy Classifier or a neural network used becomes. 11. The method according to any one of claims 1 to 10, characterized, that a single converter is used, which is so probably as an actuator for generating the excitation impulses as well as a sensor for receiving the Im pulse response serves. 12. The method according to any one of claims 1 to 10, characterized, that an actuator to generate the excitation impulses and several sensors to receive the Im pulse response can be used. 13. The method according to any one of claims 1 to 12, characterized, that the transducer (s) have contact tips with which they are in contact with the component be brought.   14. The method according to any one of claims 1 to 13, characterized, that the converter or the converter and the component without me mechanical influence on the component with a defined force are pressed against each other, which have a suitable physical effect is caused. 15. The method according to any one of claims 1 to 14, characterized, that the transducer to generate the excitation impulses controlled with an electrical pulse which is the inverse transfer function of the Has converter. 16. The method according to any one of claims 1 to 15, characterized, that the transducer matched the component like this be dimensioned that the resonance indicators of the component due to influences of the Transducers are covered. 17. The method according to any one of claims 1 to 16, characterized, that piezoelectric transducers are used. 18. The method according to any one of claims 1 to 10 and 12 to 17. characterized, that as a converter to receive the impulse response Laser scanning vibrometer is used.