WO2023041219A1

WO2023041219A1 - Evaluating rounded sections of inner bore edges by means of machine learning using vibration data or the like

Info

Publication number: WO2023041219A1
Application number: PCT/EP2022/069068
Authority: WO
Inventors: Peter Hoeh; Stephan Wengerek; Ruben KAPP; Mario Herrera
Original assignee: Robert Bosch Gmbh
Priority date: 2021-09-15
Filing date: 2022-07-08
Publication date: 2023-03-23
Also published as: CN118234589A; DE102021210205A1

Abstract

The invention relates to a method for rounding inner bore edges of a common rail component, having the steps of flushing an inner bore using an electrolyte solution or providing a flow of electrolyte solution through the inner bore during an electrochemical machining (ECM) process, detecting vibrations in the common rail components in the region of the inner bore in question using at least one vibration detection device over at least one specified period of time during the flushing process, filtering the detected vibration data in order to obtain vibration characteristics, and evaluating the vibration characteristics using at least one machine-learned classifier in order to evaluate the quality of the rounded section in the inner bore.

Description

Title:

Assess fillets of inner bore edges using machine learning using vibration data or the like

The present invention relates to a method for rounding inner bore edges and a system for evaluating rounding of inner bore edges of a common rail component, injection nozzle or the like.

State of the art

Fuel injection systems for internal combustion engines of motor vehicles, such as so-called common rail systems, include fuel-carrying components that are subjected to an internal pressure of usually up to 1600 bar. This internal pressure is still subject to strong fluctuations. Accordingly, high strength requirements are placed on such components.

For example, a component subject to internal pressure is known from EP 1 435 454 A1, which has a vertical inflow channel for connection to a high-pressure fuel accumulator and a horizontal connection channel branching off transversely therefrom for connecting an injector or the like. A horizontal connection duct and an inflow duct form an intersection area in which a cross-section results that leads to lower local stresses.

It is also known to round off sharp edges on inner bores of such components in order to prevent notch effects and consequent damage to the components. For example, an electrochemical machining process (ECM) can be used for rounding, the machined Holes are usually checked with an endoscope. In an ECM process, an electrode is inserted into the inner bore of the component and a conductive electrolyte is pumped through the inner bore. Material is removed by the action of an electric current and the rounding occurs. The ECM process usually has three phases. In a first phase, in which the inner bore is cleaned or pre-rinsed, the electrolyte is pumped through the inner bore at high pressure in order to free it from particles from the inside. In a second phase, which is the actual ECM processing, material is removed using an electric current between an electrode and the workpiece. In a subsequent, third phase, the electrolyte is pumped through the inner bore again in order to carry out cleaning.

Disclosure of Invention

The use of machine learning methods is known in order to save or reduce a complex visual inspection using an endoscope. The aim is to be able to reliably make a prediction based on process data from the ECM process as to whether the rounding of inner edges is sufficient. The process data can include current and volume flow curves. However, it has been found that the process data is occasionally incomplete or insufficient to make a reliable prediction.

It is consequently an object of the invention to propose a method for rounding the edges of the inner bore of a common rail component, in which the quality of the rounding can be better assessed on the basis of process data.

This object is solved by a method having the features of independent claim 1 . Advantageous embodiments and developments can be found in the dependent claims and the following description. There is a method for rounding inner bore edges of a common rail component, comprising the steps of flushing an inner bore with an electrolyte solution during electrochemical machining (ECM), detecting vibrations and/or structure-borne noise in the common rail component in the area of the relevant Inner bore using at least one vibration detection device for at least a predetermined period of time during flushing, filtering of detected vibration data to obtain vibration characteristics using a transient signal, a Fourier transformation, autocorrelations, wavelet transformation or the like and evaluating these vibration characteristics using at least one machine-taught classifier to evaluate the quality of the fillet in the internal hole.

A common rail component can in particular include a high-pressure accumulator, which is also referred to as a fuel distributor pipe or “rail”. Injectors can be connected to the high-pressure accumulator via high-pressure connections, which have a flow direction running transversely to a main extension direction of the high-pressure accumulator, and are supplied with fuel from the high-pressure accumulator. Other components in the form of valves or injector bodies are conceivable.

A classifier can be understood as an algorithm for automatically carrying out a classification method. The classifier could be trained by a machine learning process to distinguish between at least two different classes, which in the present case could be understood as “not sufficiently rounded” and “sufficiently rounded”. The training can take place, for example, by means of supervised learning while the classifier is in operation.

The method according to the invention is generally based on a conventional ECM method. This means that the three processing phases that are customary in the prior art and are described above are carried out. After a pre-rinse, the actual machining with material removal is carried out and a post-rinse process follows to clean the relevant inner bore. However, one focus here is on the Evaluation of the quality of the fillet in the inner bore. For this purpose, flow and volume flow curves during the ECM processing, in particular in the processing phase or the post-rinsing phase, can also be used, as is customary in the prior art. However, it is particularly advantageous to detect vibrations in the common rail component in the area of the relevant inner bore by means of at least one vibration detection device.

If the inner bore in question is not properly rounded, the liquid flowing in it is subjected to severe deflection, which leads to the formation of turbulence within the inner bore. This can cause vibrations that stimulate the relevant common rail component to oscillate. These can be recorded and taken into account when evaluating the quality of the fillet. The detection can be carried out by means of one or more vibration detection devices. The delivered vibration data can be filtered in a suitable way in order to enable an evaluation. As a result, at least one further classifier is made available to an already existing machine learning process, which enables the rounding to be evaluated. By considering vibration characteristics when evaluating the quality of the rounding, which is completely unknown in the prior art, a significantly higher level of precision can be achieved.

The at least one vibration detection device could be attached to the common rail component and/or to an ECM electrode. It is particularly useful to arrange the corresponding measuring points as close as possible to the inner bore in question. For example, a vibration detection device could be arranged directly on the high-pressure accumulator. A central measuring point in the high-pressure accumulator could be advantageous. If an arrangement on the high-pressure accumulator cannot be carried out when executing the ECM method, the distance to the high-pressure accumulator should be as small as possible and in particular should not exceed 15 cm.

The at least one vibration detection device can include a laser vibrometer, so that vibration data is detected without contact on at least one section of the common rail component. a laser Vibrometers may include a laser light source that is aimed at the surface of the common rail component. If the surface moves due to vibration, the frequency of the backscattered laser light shifts due to the Doppler effect and can be evaluated to obtain frequency and amplitude. It can be particularly useful to aim the laser vibrometer at several sections of the component in question one after the other in order to generate a more complete set of data about the vibration behavior of the component. Using a laser vibrometer increases flexibility in assessing fillet quality and does not require mechanical interaction with the component in question.

However, it is also conceivable that the at least one vibration detection device includes a piezo or acceleration sensor that is mechanically connected to the common rail component. This could reduce the cost of the vibration detector and the component could have a dedicated flange or other prepared or marked surface portion to which the vibration detector could be attached for fillet quality evaluation. The at least one vibration detection device can, for example, be glued on and completely detached again after the end of the method.

It is advantageous if vibration features are extracted from transient vibration data by a Fourier transformation and/or by a wavelet transformation. In this embodiment, the filtering step therefore includes a Fourier transformation, an autocorrelation and/or a wavelet transformation. In particular, the wavelet transformation is able to achieve improved resolution at different vibration frequencies, for example by having better frequency resolution at lower vibration frequencies and better time resolution at higher vibration frequencies. The individual measurement data can be subdivided into short sections following one another in terms of time and by the Fourier transformation, in particular a short-time Fourier transformation (STFT) or a fast Fourier transformation or a wavelet transformation are processed. The individual sections can have a length of about 1 to 10 s. Through the processing, significant vibrational features can be obtained in the time and/or frequency domain, which are suitable as classifiers for machine learning.

It is nevertheless conceivable that vibration features are additionally extracted directly from transient vibration data. These can include vibration amplitudes, for example.

The evaluation of the vibration characteristics could additionally be carried out using at least one further machine-taught classifier, with the further classifier being based on flow or volume flow curves during flushing and/or in a preceding ECM processing phase. The evaluation of a rounding quality using a method known in the prior art and based on machine learning can therefore be expanded and supported by additional classifiers that result from the detected vibrations.

Analogously to the previous statements, the invention also relates to a system for evaluating rounding of inner bore edges of a common rail component, having a computing unit and at least one vibration detection device coupled to the computing unit. According to the invention, it is provided that the computing unit is designed to detect vibrations in the common rail component in the region of the relevant inner bore by means of the at least one vibration detection device for at least a predetermined period of time, in particular when flushing an inner bore with an electrolyte solution during electrochemical machining (ECM). , filtering acquired vibration data to obtain vibration characteristics, and evaluating vibration characteristics using at least one machine-taught classifier to assess the quality of the fillet in the internal bore.

The at least one vibration detection device could include a laser vibrometer, so that vibration data is detected without contact on at least one section of the common rail component. The arithmetic unit could also be designed to evaluate the vibration characteristics using at least one additional machine-taught classifier, the additional classifier being based on current or volume flow profiles during flushing and/or in a preceding ECM processing phase.

Further measures improving the invention are presented in more detail below together with the description of the preferred exemplary embodiments of the invention with the aid of figures.

exemplary embodiments

It shows:

FIG. 1 shows a schematic view of a common rail component;

FIG. 2 shows a comparison of the inner edges of a bore with and without rounding;

FIG. 3 shows a schematic, block-based representation of a method according to the invention;

FIG. 4 shows a diagram with volume flow and vibration data; and

Figure 5 is a schematic view of vibration characteristics.

FIG. 1 shows a common rail component in the form of a high-pressure accumulator of a common rail system. A plurality of connections 4 are provided on the component 2 , each of which has a bore 6 which extends into the interior of the component 2 transversely to a main direction of extent x of the component 2 . They each end in a throttle bore 8, which is shown in an enlarged sectional view. Inner bore edges 10 are present on the throttle bore 8 and are to be rounded off by an ECM process are. For this purpose, an electrode 12 is provided, which is placed in the component 2 . Holes 6 and 8 are flushed with an electrolyte solution during the ECM process. If a voltage is present between the electrode 12 and the component 2, material is removed from the surfaces around which the flow occurs by means of the electrolyte solution.

A vibration detection device 13 is arranged on a surface of the component 2 and is designed to detect a vibration on the surface. The vibration detection device 13 is connected to a computing unit 11, which receives or detects the detected vibration data and can consequently use it to carry out the method described further below. The vibration detection device 13 and the computing unit 11 can consequently represent a system for evaluating rounding of inner bore edges 10 .

FIG. 2 shows in a left-hand part of the plane of the drawing that inner bore edges 10 that are not rounded can cause turbulence when a liquid, for example fuel, flows through. In a right-hand part of the plane of the drawing, the inner bore edges 10 are each rounded, so that the flow is in contact with the surfaces of the bores 6 and 8 . It makes sense to check the rounding of the inner bore edges 10 to ensure an even, harmonious flow. This is accomplished by a method illustrated in FIG.

The method for rounding the inner bore edges 10 of the common rail component 2 includes the step of rinsing 14 the inner bore 8 with an electrolytic solution during electrochemical machining (ECM). In particular, this can be a first phase or a third phase of an ECM process, since this involves working with significantly high pressure and flow. In a first flushing process without electrochemical removal, only the electrolyte solution is flushed through the inner bore 8 in order to clean it. This is followed by a second phase in which material is removed by means of the electrolyte solution using an electric current. In a third phase, rinsing is carried out again without electrochemical removal in order to achieve post-cleaning. It is particularly advantageous if the Step of rinsing 14 of the method according to the invention corresponds to the third phase. However, it can also be useful to carry out an ECM processing operation several times and to carry out the method according to the invention at least partially during the first phase.

During flushing 14 or directly during ECM processing, vibrations in common rail component 2 in the region of relevant inner bore 8 are detected 16 by at least one vibration detection device 13 over at least a predetermined period of time. Detected vibration data is then filtered 18 to obtain vibration characteristics. Finally, the vibration characteristics are evaluated 20 using at least one machine-taught classifier in order to assess the quality of the rounding in the inner bore 8 . If it is determined that the rounding is not yet sufficient, the ECM processing can be carried out again 22 in order to then check the rounding again.

In a diagram, FIG. 4 shows an example of a volume flow curve (upper part of the diagram) and a vibration curve (lower part of the diagram) in three phases of an ECM processing method. A first phase 24, the rinsing phase, is followed by a second phase 26, the material removal phase, and a third phase 28, the post-rinsing phase. The first phase 24 and the third phase 28 are characterized by a very high volume flow and comparatively low vibration. The second phase 26, in which material is removed, has a very low volume flow, but stronger vibrations. According to the prior art, among other things, the volume flow data from the second phase can be used to assess rounding. According to the invention, however, vibration data from one of the two flushing phases 24 or 28 is used in order to carry out an evaluation of the rounding based on machine learning. As can be seen in a detail shown in the left part of the plane of the drawing, a short period of time, for example 2 s, can be used. After filtering 18, vibration features 30 can be examined. Finally, FIG. 5 shows, by way of example, various vibration characteristics 30 which, for example, show different amplitudes, smallest, mean, average, maximum frequencies and the like. As mentioned above, mathematical features/characteristics are extracted from Fourier transform, wavelet transform, autocorrelation and others

Characteristics, for example, assigned certain coefficients from the autocorrelation, number of peaks or the like. From around 1500 extracted features, 20-30 features are then usually used in practice.

Claims

Expectations

1. A method for rounding inner bore edges (10) of a common rail component (2), comprising the steps:

Flushing (14) or flowing through an inner bore (8) with an electrolyte solution in electrochemical machining (ECM), characterized by

Detection (16) of vibrations in the common rail component (2) in the region of the relevant inner bore (8) by means of at least one vibration detection device over at least a predetermined period of time during flushing (14) or throughflow, and

filtering (18) detected vibration data to obtain vibration characteristics (30),

Evaluating (20) the vibration characteristics (30) using at least one machine-taught classifier in order to evaluate the quality of the rounding in the inner bore (8).

2. The method according to claim 1, characterized in that the at least one

Vibration detection device on the common rail component (2) and / or on an ECM electrode (12) is attached.

3. The method according to claim 1 or 2, characterized in that the at least one

Vibration detection device comprises a laser vibrometer, so that the detection (16) of vibration data is contactless on at least a portion of the common rail component (2).

4. The method according to any one of the preceding claims, characterized in that the at least one

Vibration detection device comprises a piezo or acceleration sensor, which is mechanically connected to the common rail component (2).

5. The method according to any one of the preceding claims, characterized in that vibration features (30) are extracted from transient vibration data by a Fourier transformation and/or by a wavelet transformation.

6. The method according to any one of the preceding claims, characterized in that vibration features (30) are additionally extracted directly from transient vibration data.

7. The method according to any one of the preceding claims, characterized in that the evaluation (20) of the vibration characteristics (30) is additionally carried out using at least one further machine-taught classifier, the further classifier being based on current or volume flow profiles during the flushing (14) and/or in a previous ECM processing phase (24, 26, 28).

8. System for evaluating fillets of inner hole edges

(10) a common rail component (2), having: a computing unit (11) and at least one vibration detection device (13) coupled to the computing unit (11), characterized in that the computing unit (11) is designed to do so during flushing (14) or an electrolytic solution flowing through an inner bore (10) during electrochemical machining (ECM) Vibrations in the common rail component (2) in the area of the relevant inner bore (10) by means of the at least one vibration detection device (13) over at least one predetermined detecting period of time, filtering (18) detected vibration data to obtain vibration characteristics (30), and Evaluate (20) vibration characteristics (30) using at least one machine-taught classifier in order to evaluate the quality of the rounding in the inner bore (8).

9. System according to claim 8, characterized in that the at least one vibration detection device (13) comprises a laser vibrometer, so that the detection (16) of vibration data takes place without contact on at least one section of the common rail component (2).

10. System according to claim 8 or 9, characterized in that the computing unit (11) is further designed to evaluate (20) the vibration characteristics (30) additionally using at least one other machine-taught classifier, the other classifier based on current or

Volume flow curves during the flushing (14) and / or in a previous ECM processing phase (24, 26, 28) is based.