WO2017051130A1

WO2017051130A1 - Method for detecting manufacturing defects using the acoustic emission technique on a stamping tool, and associated device

Info

Publication number: WO2017051130A1
Application number: PCT/FR2016/052411
Authority: WO
Inventors: Pierre Friedrich; Samuel MAILLARD
Original assignee: Centre Technique Des Industries Mecaniques
Priority date: 2015-09-25
Filing date: 2016-09-23
Publication date: 2017-03-30
Also published as: FR3041749A1; FR3041749B1

Abstract

A method for monitoring a tool for shaping by stamping, comprising the following steps: acquiring at least one signal (4) recorded, in the form of an acoustic amplitude as a function of time, by a sensor placed on the tool, over at least one length of time (Δt) during which a part in the process of being manufactured is compressed by the tool or is compressed against the tool; a step of temporal analysis of the recorded signal (4), in order to calculate at least one mathematical signature (S_p) generated by a statistical type calculation on the series of acoustic amplitudes (X_n); calculating a coherence criterion (Crit_p) depending on the mathematical signature of the part (S_p) and depending on at least one mathematical signature (S_{p - 1}, S_p _-2, S_p-3, S_{p -4}) of at least one other part belonging to a varying group of parts shaped beforehand on the same tool; comparing the value (Crit_p) of the coherence criterion with at least one threshold (S).

Description

METHOD OF DETECTING MANUFACTURING DEFECTS USING ACOUSTIC TRANSMISSION TECHNIQUE ON A

TOOL FOR HITTING, AND ASSOCIATED DEVICE.

The object of the invention is to monitor processes for shaping workpieces on forming tools by striking, or more generally by compressing a metal sheet or blank, which is generally made of metal. Such monitoring is intended to prevent the manufacture of defective parts and / or to prevent damage to the tool. Industrial process monitoring systems are known. They use, for example, vibratory phenomena at the tool, with a frequency greater than 1 6 kHz. The invention is in the field of monitoring methods using more particularly acoustic emission, which typically corresponds to frequency ranges between 95 and 950 kHz. Frequency filtering may be further applied to further restrict the range of signals recorded within this range.

The compression of the blank or the sheet by the tool, itself generally actuated by a press, causes a vibration of the tool which depends in particular on the forces undergone by the tool, and therefore the process of deformation of the part, the process of deformation having repercussions on the mechanical interactions between the part and the tool. When phenomena of mechanical instability occur at the part and / or the tool, for example by occurrences of local metallurgical instabilities of crack, crack or plastic instability type, the interaction of the piece and the The tool is modified, and the vibratory diagram of the tool is also modified. It can therefore be expected that the signals emitted, in certain well-chosen frequency ranges, could make it possible to detect the appearance of such instabilities during the formatting process. For example, it has been proposed to use frequency analyzes of the transmitted signals to detect the presence of unusual frequencies with respect to a predefined "normal" frequency range. Such methods, however, are expensive in calculation, and abnormal frequency domains may be difficult to delimit.

The object of the invention is to propose a process for monitoring the manufacture of a formatting tool, which is less expensive in calculation means, and which is easily calibrated according to the tool to be monitored and the part to be measured. produce on this tool.

To this end, the invention proposes a method of monitoring the manufacture of a batch of parts stamped by a striking tool, the method comprising the following steps:

an acquisition of at least one recorded signal, in the form of an acoustic amplitude as a function of time, by one or more piezoelectric sensor (s) placed on the tool, or placed on a matrix supporting the coin struck by the tool

a step of temporal analysis of the recorded signal, for calculating from the series of acoustic amplitudes composing this recorded signal, at least one mathematical signature in the form of a scalar value associated with said piece, and resulting from a calculation of type statistic on the sequence of acoustic amplitudes, -a calculation of a coherence criterion value as a function of the mathematical signature of the part and as a function of at least one mathematical signature of at least one other piece belonging to a group evolution of parts previously formatted on the same tool,

a comparison of the value of the coherence criterion with at least one threshold.

the emission of an alert if the criterion of coherence exceeds the threshold,

for each piece for which a coherence criterion value has been calculated, a step of redefinition of the group evolution of parts, depending on the value of the consistency criterion.

Acquisition of the recorded signal is done over a time interval that includes at least one time range during which a workpiece being manufactured is compressed by the tool or is pressed against the tool. The recording of the piezoelectric sensor can be triggered by a sensor able to detect a particular position and / or able to detect a particular movement of the tool.

By sensor placed on the tool is meant a sensor placed in contact with the tool or placed in sufficient proximity to the tool to receive acoustic emissions specific to the operation of the tool.

According to a preferred embodiment, a new selection of the said at least one other piece whose mathematical signature is used for the calculation of the criterion value, is performed with each new shaped part. The result of the selection may however sometimes lead to keeping the selection of the same at least one other piece, if the criterion calculated for the part being manufactured exceeds the threshold.

Advantageously, if for a given piece, the criterion does not exceed the threshold, the mathematical signature of this piece is then used to calculate the consistency criterion of at least one other piece subsequently shaped on the same tool. The last pieces of a production batch can of course be an exception to this rule.

In other words, the reference value or values are calculated from an evolutionary group of previous mathematical signatures, the evolutionary group thus retained being redefined from one piece to the next.

However, if the consistency criterion exceeds the threshold for a given part, the same group of mathematical signatures can be retained to calculate the consistency criterion of the next part formatted on the same tool, in order not to include outliers in the group of mathematical signatures.

According to the variant embodiments, the alert can be triggered by exceeding a threshold by a higher value, by exceeding a threshold by a lower value, or by exceeding a threshold in absolute value.

By issuing an alert, it is understood, for example, to deliver an audible message, and / or to deliver a message on a screen, and / or to enter a specific value in a production tracking file making it possible to associate this alert with the piece set. form.

The alert can then be interpreted as indicating that there is a higher than average probability that the part has a production defect, and the part can be controlled / can be scrapped.

The alert can also be interpreted as a need to change the settings of the production tool.

By temporal analysis is meant an analysis of the values of the signal as a function of its variations over time.

Such an analysis may be preceded if necessary by one or more intermediate filterings, to obtain a filtered signal which remains a signal as a function of time, without having to convert the signal, of the time domain in which it has been recorded. to a frequency domain.

By acoustic signal is meant a signal whose frequency range extends over frequency values typically between 95 kHz and 950 kHz.

The sensor (s) used may be, for example, piezoelectric type sensors. The sensor or sensors used may be, for example, piezoelectric type sensors whose resonance frequency is around

200kHz.

The recorded signal can be the object of a first temporal filtering in order to eliminate a part of the background noise of the signal. By mathematical signature of statistical type is meant a scalar value obtained by a function, or by a combination of functions, usually used in statistics to characterize a distribution of values, the distribution of values being here represented by the succession of acoustic amplitude values. composing the recorded signal.

According to a preferred embodiment, the calculation of the coherence criterion is carried out taking into account the mathematical signature of the piece, as well as a predefined number of mathematical signatures computed on an evolutive group of pieces previously formatted.

By evolutionary group is meant a group whose composition varies. For example, the group of pieces can be defined as the group of last pieces whose consistency criterion has not exceeded the threshold, by taking a number of pieces corresponding to the predefined number.

Advantageously, the predefined number is greater than or equal to two, and preferably greater than or equal to 5, advantageously greater than or equal to 7. The previously formatted parts are typically the last parts that are passed on the production tool. before the current part, excluding values associated with parts for which the consistency criterion has exceeded the threshold.

The threshold can be exceeded by high value or low value, depending on how the criterion is defined.

According to one embodiment, the coherence criterion takes into account a series of coherence values, in the statistical sense of the term, each coherence value translating a relationship of difference between the signal recorded on the part being manufactured, and a recorded signal of a piece of the evolving group of pieces previously produced on the tool.

Preferably, the evolving group of parts includes parts all belonging to the same production batch as the part being manufactured. According to one embodiment, at the start of a new production batch, the first values of mathematical signatures retained for calculating reference values can be selected manually by an operator.

Thus the operator can verify that the mathematical signatures selected actually correspond to visually compliant parts or from the point of view of a control mode independent of the acoustic measurements made on the tool.

Advantageously, the coherence criterion can take into account a between the mathematical signature of the piece and an average of mathematical signatures of a constant number of pieces among the evolutive group of pieces previously formatted on the tool. Advantageously, the group of parts whose mathematical signatures are used for calculating the consistency criterion changes with each new piece produced on the tool.

The average can be an arithmetic mean, an algebraic average, or some other type of average. According to yet another variant embodiment, the coherence criterion can take into account a difference between the mathematical signature of the piece and an average of mathematical signatures calculated on an increasing number of pieces previously shaped on the tool. The average may for example be an average calculated on all the parts of the production batch produced on the same tool, and judged to be compliant either by means of the process itself, or by a visual analysis, or by another mode of control.

In some embodiments, the consistency criterion takes into account an average of mathematical signatures.

In other embodiments, the coherence criterion takes into account, for each of the pieces of the evolutive group of pieces previously formatted on the tool, at least one relative difference in values between the signature. the mathematical signature of the workpiece being manufactured, and the mathematical signature of said each of the pieces of the evolutive group of pieces.

According to one embodiment, the relative differences of values are reported each time to the value of the mathematical signature of the piece, that is to say that the relative difference is the difference between the two mathematical signatures divided by the signature. mathematical of the piece.

According to another embodiment, the relative differences in values are each referred to the value of the mathematical signature of the piece previously formatted, that is to say that the relative difference is the difference between the two signatures. mathematics divided by the mathematical signature of the piece previously formatted.

According to yet another embodiment, the criterion of coherence takes into account, for each of the parts of the previously formatted group of pieces, both a first relative difference related to the mathematical signature of the piece, and a second relative difference. relative to the mathematical signature of the piece of the group of previously formatted pieces.

According to an alternative embodiment, the criterion can take into account absolute differences in values between the mathematical signature of the part being manufactured and the mathematical signatures of parts previously produced on the same tool.

According to a variant embodiment, the coherence criterion can take into account a sum, or a sum of absolute values, of the relative differences in value between the mathematical signature of the piece and each of the mathematical signatures of the pieces of the evolutive group of pieces set in motion. form previously on the tool.

According to another variant embodiment, the criterion of coherence can take into account a sum -or a sum absolute values-, discretized values of the relative differences in values between the mathematical signature of the piece and each of the mathematical signatures of the pieces of the evolutive group of pieces previously formatted on the tool.

By discretizing an initial sequence of values, it is intended to replace each of the values of the initial sequence by a value forming part of a predefined sequence, as a function of the positions of each of the initial values relative to the values of the predefined sequence.

The relative differences of values of mathematical signatures can thus be replaced by values corresponding to a limited number of steps, according to an increasing monotonic function or according to a decreasing monotonic function.

The sum, or the sum in absolute value, then adds terms corresponding to these steps. The values of the steps used to calculate the consistency criterion can thus be easily visualized in a graphical manner, and an operator can more easily judge the stability of the manufacturing process of the part.

The display, for example to the attention of an operator monitoring the manufacture of the suite of rooms, can be done for example in the form of a matrix of patterns of gray levels or color intensity corresponding to the different steps .

The matrix may comprise, for example, a succession of lines, a line showing the values of the steps calculated for the last piece formatted. The columns of the matrix then correspond to the different comparisons made between a manufactured part, and the parts that precede it in the production order.

For example, an array matrix can be displayed, each box displaying a pattern, a gray level or a color intensity, which translates the discretized value of a relative difference of mathematical signature values, by a list of predefined correspondence linking the patterns to the different levels allowed for the discretized values.

The table may for example comprise a succession of lines, a new line being added to the table to display the values of the calculated steps for the last piece formatted.

A column of the table then corresponds, in discretized values, to a series of relative differences of mathematical signatures, calculated according to the same formula, and performed between, each time, a fabricated part, and a part which precedes it by an order " n "characteristic of the column, in the order of manufacture.

According to a preferred embodiment, the recorded signal is a sequence of sampled values X _n , and the mathematical signature of the part is of the type:

Where q is the number of successive values X _n of the signal recorded during the coin strike, k and μ are either constant values or values themselves calculated according to the sequence of values X _n , a and β are whole or non-integer exponents that remain the same for all formatted parts.

For example μ can be is an average value of these successive values X _n , k can be equal to the inverse of the standard deviation σ of the sequence of values X _n , a can be equal to 4 and β can be equal to 1. S (piece) is then equal to a value also called "kurtosis" of the series of values representing the acoustic signal

According to a preferred embodiment,

k = 1, μ = 0, α = 2 and β = 0.5 S _p is then equal to a value commonly known as "squared mean" of the sequence of values representing the acoustic signal.

For example, the recorded signal is a sequence of time-sampled values X _n , and the mathematical signature of the part is a root mean square of the sequence of values forming the recorded signal.

According to another embodiment, the recorded signal is a sequence of time-sampled values X _n , and the mathematical signature of the piece is a kurtosis value of the series of values forming the recorded signal.

Advantageously, the beginning of the recording time range can be calculated from a synchronization signal representative of the position of a movable portion of the tool during the forming of the workpiece.

The signal representative of the position of the tool may be for example a cam signal, that is to say a signal triggered by the contact of a cam with the tool.

According to one embodiment, the recorded signal is acquired over a continuous time range.

According to another embodiment, recorded signals are acquired over several non-contiguous time slots for each punch of the coin.

According to one embodiment, a single mathematical signature value is calculated during the passage of a given piece on the tool.

According to another embodiment, several mathematical signature values are calculated during the passage of a given piece on the tool.

The different mathematical signature values may correspond to different statistical functions and / or may correspond to the application of these statistical functions to different time ranges on which an acoustic signal has been recorded during the production of the same part. The triggering of the start of the recording does not necessarily occur at the moment of the synchronization signal, but may for example be clocked according to a sequence of synchronization signals.

According to an advantageous embodiment, a series of graphic elements representative of discretized values of several of the last calculated mathematical signature values can be displayed simultaneously on a screen.

According to variant embodiments, the displayed graphic elements may be representative of discretized values of a series of two-by-two differences of recent calculated mathematical signature values, or of a series of relative differences two by two, of recent mathematical signature values. calculated.

For example, it is possible to display on a screen, or to print on a tracking sheet, a matrix of rectangles whose gray or luminosity levels represent the discretized value, in order to simply visualize the variability of the last calculated statistical characteristic values. In order to simply visualize the variability of the last calculated mathematical signatures, we can display on a screen, or print on a tracking sheet, an array matrix whose cells display different colors, different levels of gray or different levels of brightness, each level. representing a discretized value of a statistical type value whose evolution is to be followed visually, the value of statistical type being calculated from at least two mathematical signatures of a series of parts being manufactured on the tool .

The matrix may comprise several rows or several columns, a row or a column being each time associated with a statistical characteristic value of a given type. The invention also proposes a device for monitoring the quality of parts shaped by striking in a tool, comprising:

at least one acoustic emission sensor, configured to acquire a recorded signal comprising a series of acoustic amplitudes as a function of time,

a synchronization unit configured to trigger, from a time calculated according to a signal or a series of synchronization signals, the acquisition of a series of acoustic amplitudes by the sensor for a given duration, to obtain a recorded signal of the temporal type,

a statistical processing unit configured to calculate a mathematical signature of the recorded signal, in the form of a scalar value associated with a part being manufactured, by a statistical type calculation on the series of acoustic amplitudes,

a tracking memory configured to keep in memory an evolutionary sequence of one or more previously calculated mathematical signatures for an evolving group of pieces previously formatted on the tool, or for storing a sliding average of previously calculated mathematical signatures

A criterion calculation unit configured to calculate a coherence criterion value according to the mathematical signature associated with the piece, and according to at least one mathematical signature associated with another piece previously produced on the same tool, and

An alert unit configured to compare the consistency criterion with at least one threshold, and configured to, if the consistency criterion exceeds the threshold, trigger an alert, and configured to, if the coherence criterion remains below the threshold, record in the tracking memory, the mathematical signature calculated for the part being manufactured, or modify a value in the tracking memory according to the mathematical signature calculated for the part being manufactured.

The same tracking device can be used to implement the monitoring method according to the invention on different tools, and / or for the production of different parts on the same tool, by means of an initial calibration phase of the process, which can to be done simply by validating the mathematical signature records of a predefined number of first pieces "compliant" on the tool, the conformity being judged independently of the values recorded for the needs of the process.

Some objects, features and advantages of the invention will become apparent on reading the following description, given solely by way of nonlimiting example, and with reference to the appended figures, in which:

FIG 1 is a schematic representation of a manufacturing tracking device according to the invention, and

FIG. 2 is a simplified monitoring unit operating algorithm belonging to the device of FIG. 1.

As shown in FIG. 1, a monitoring device 1 intended to be mounted on a shaping tool 10, comprises a monitoring unit 7, and comprises an acoustic emission sensor 2, which can be associated with a acoustic filter 3, the sensor delivering an acoustic signal 4 recorded by the monitoring unit 7. The monitoring device 1 also comprises a synchronization device 5 for delivering to the monitoring unit 7 a signal 6, for example a signal cam corresponding to a particular position of a cam 1 1 imposing the movement of a movable portion 10b of the shaping tool 10. The synchronization device 5 thus makes it possible to locate in time the successive instants of the setting in the form of a particular piece 20 _p on the tool 10, the index p representing for example the pth piece of the current production batch. The monitoring unit 7 is configured to enable to control the formatting tool 10 or at least to send to a control unit of this tool 10 a signal 9 to pause. This pause signal 9 is sent if a criterion calculated from the characteristics of the recorded signal 4 exceeds a certain threshold.

For each piece produced, the monitoring unit 7 is configured to use a recorded signal range 4 corresponding to a time range of predefined length and whose position in time is determined from the synchronization signal 6. The unit 7 comprises a man-machine interface allowing an operator to set certain parameters for recording and / or taking into account the recorded signal 4, for example allowing the operator to position the time window on which the recorded signal 4 is analysis.

The interface may also be configured to allow the monitoring unit 7 to deliver an alert message 8 to the attention of the operator, for example in the form of an auditory message or a visual message. Such a signal can be sent to the operator at the same time as the monitoring unit 7 sends the pause signal 9 to the tool 10.

The tool 1 0 typically comprises a die 10a and a movable portion or a punch 10b movable relative to the die. Blanks or blanks of material, for example sheet metal blanks are placed successively between the punch and the die to be shaped in the same geometry.

In the example illustrated in FIG. 1, a sheet metal coil 13 is thus unrolled between the punch and the die in order to print a specific geometry by plastic deformation.

The tool can further be configured to cut out a contour or part of the contour of the successive pieces produced.

The acoustic emission sensor 2 is typically placed in contact with a punch element or a die element of the tool. Preferably, the sensor may be placed in contact with the matrix which remains stationary.

According to other embodiments, the sensor can be inserted into a recess of the tool, the recess being arranged so as to place the acoustic emission sensor near the "sensitive" areas of the tool, that is to say areas of the tool printing plastic deformations amplitude or gradient may give rise to defects on the part.

Acoustic emission sensors, synchronization devices and filters for eliminating background noise are known and will not be described here.

FIG. 2 schematically illustrates the operating principle of a monitoring unit 7 according to the invention, such as the monitoring unit of FIG. 1. FIG. 2 shows some common references to FIG. same references designating the same elements.

The monitoring unit 17 comprises a synchronization unit 14, a buffer memory 15, a statistical processing unit 16, a tracking memory 17, a criterion calculation unit 18, and an alert unit 19.

As illustrated in FIG. 2, a synchronization unit 14 of the monitoring unit 7 receives the synchronization signal 6, for example in the form of a square signal passing from 0 to 1 at a time t ₀ . According to an alternative embodiment, the monitoring unit 7 can, from this initial moment t ₀ , make successive acquisitions at times t _n spaced evenly spaced, and distributed over a total time interval Δΐ. Successive acquisitions at times t _n are at a frequency known as the sampling rate, which can typically be between 1 kHz and 1 MHz. Multiple synchronization modes are of course possible. The monitoring unit 7 may, for example, acquire all the acoustic signals available between two instants T ₀ and T _15, the instants TQ and Ύι being distant from the initial instant t ₀ and defined in relation to this one. The initial time t ₀ can be used to set an initial start time T ₀ for recording start, and the monitoring unit 7 can be configured to then record a predefined number "n" of acoustic amplitude values of the signal. More generally, the monitoring unit 7 comprises a synchronization unit 14 which, starting from the reception of the synchronization signal 6, triggers the acquisition in a buffer memory 15 of a sequence Xi X ₂ . X _n of acoustic amplitude values taken from the signal reception 4 from the sensor 2, after a first possible filtering.

The acoustic amplitude values acquired in the buffer memory 15 and associated with the part 20 _p currently being shaped on the tool, whose associated values are indicated by an index "p" in FIG. 2, are sent to a statistical processing unit 16 of the monitoring unit 7. The statistical processing unit 16 calculates, from the acoustic amplitude sequence recorded in the memory 15, one or more mathematical signatures S _p associated with the _p- th piece produced on the tool.

The monitoring unit 7 comprises a criterion calculation unit 18. The criterion calculation unit 18 is configured to calculate at least one reference value taking into account the previously calculated mathematical signatures for other identical shaped pieces. on the same tool. The reference value may be for example an average of mathematical signatures of parts produced before the part being manufactured. According to one embodiment, the reference value may be an average of the mathematical signatures of all the pieces previously produced in the same production batch and found to be compliant. In another embodiment, the reference value may be a running average of a constant number of mathematical signatures of pieces previously produced in the same production lot and found to be compliant. According to yet another embodiment, the reference values retained for the current part may be a group of values comprising a predefined number of mathematical signatures of parts previously produced in the same production batch and found to be in conformity.

The criterion calculation unit 1 8 is furthermore configured to calculate at least one criterion value of coherence "Crit _p " associated with the part "p" being formatted, starting from the mathematical signature (s). _p of the part currently formatted on the tool, and the reference value or values calculated taking into account the previously acquired signatures for the other parts.

The one or more criterion values are then sent to an alert unit 1 9 which compares the one or more criterion values with one or more "S" threshold values. The threshold value (s) remain constant during the monitoring of the same production batch. If the criterion value or values are all positioned on the "compliant" side of their respective threshold, the warning unit assigns the piece p a first value, for example in the form of a boolean "bool _p " indicating that the piece is compliant. The mathematical signature (s) S _p acquired for the workpiece in process then serve to update the contents of a tracking memory 17.

The content of the tracking memory 17 is then modified, for example, by adding the last calculated mathematical signature or signatures to a list of the mathematical signatures stored in the memory 17.

The mathematical signature (s) S _p acquired for the part being manufactured are then recorded in the tracking memory 17, preferably in conjunction with at least one other mathematical signature of at least one other piece previously produced. Another mathematical signature, for example the oldest mathematical signature, can be at that moment erased from memory. According to another embodiment, the content of the tracking memory 17 is modified, for example, by modifying, as a function of the last mathematical signature or signatures computed for the part being manufactured, one or more sliding averages of mathematical signatures recorded in memory 17.

If at least one criterion value is set to the "non-compliant" side of the corresponding threshold, the alert unit assigns the coin 20 _p a second value, for example in the form of a boolean "bool _p " indicating that the piece is non-compliant. The warning unit 19 can then trigger the sending of the pause signal 9 of the tool and / or the display 8 of a message to the attention of the operator. Other modalities of warning signals are of course possible.

In this way, the mathematical signature of each piece is compared to the mathematical signatures of the last pieces produced and found to be in conformity. The monitoring unit 7 can thus detect an anomaly in the course of the shaping of a part with respect to the unfolding of the previous parts. The criterion of normality of the signatures of successive parts can thus be adapted to continuous variations of the manufacturing conditions of the part, such as the ramping up of the tool, possible changes in the temperature of the tool during operation. the journal and / or any progressive changes in the mechanical characteristics of the shaped material, from one end to the other of the unrolled sheet and / or the set of preforms.

The tracking memory 17 may not contain mathematical signature values at the time when a new piece production is started on the tool: it can be configured to store a predefined number of previous part mathematical signatures as and when that the tool formats the first pieces and that the operator validates them as compliant, or the device itself validates them as conform, for example from arbitrary values pre-stored in the memory 1 7.

We will now return to some practical points of the process obj and the invention. The use of a synchronization unit 14 makes it possible to record in the acquisition memory 1 5 the corresponding acoustic amplitude values s Xi for each piece, at a given moment, at the same stage of interaction with the tool. . By limiting the acquisition of the signal to the times during which the most significant events and / or those most likely to generate manufacturing defects occur, there is a limit to the need for inappropriate digital filtering, which could obscure a part of the information contained in the initial acoustic signal.

Depending on the method to be monitored, the time interval Δΐ during which the acoustic amplitudes are recorded s, may cover the entire duration of passage of the workpiece on the tool, or may be limited to only a fraction of the time required for placing the workpiece. in the form of the obj and on the tool. The monitoring unit 7 may comprise several acoustic signal acquisition channels making it possible to simultaneously record signals from different acoustic emission sensors, either over the same time interval or over time intervals calculated from Each signal is known from a particular sensor can then be processed separately to extract a mathematical signature associated with both said sensor and the part being manufactured. The criterion calculation unit 1 8 may then be configured to independently calculate a criterion for each acoustic emission sensor, or may, according to other embodiments, be configured to calculate a multi-sensor criterion involving in the same criterion at least one mathematical signature of the production part associated with each of the sensors used. The various sensors can be placed in different places of the tool when the production of a part is done in several passes on the same tool. The monitoring unit 7 can be configured to apply the method separately to each of the passes on the tool, with a sensor dedicated to each pass.

We will then describe an example of implementation of a monitoring method using the monitoring unit 7, in a particular case where the monitoring device uses only one sensor. This method can of course be generalized to processes using several sensors.

The statistical processing unit 16 calculates a mathematical signature from the successive acoustic amplitudes of the signals recorded by the sensors, preferably in the form of a usual statistical function applied to the successive acoustic amplitudes, for example the function usually referred to in statistics as " kurtosis "or the function usually referred to in statistics as" MS ". More generally, the mathematical signatures can be calculated, with a possible constant proportionality factor, as a sum of the deviation of each acoustic amplitude to an average acoustic amplitude-average acoustic amplitude calculated on the acquired signal for the first time. part being manufactured, each term of the sum being raised to the same predefined power. The sum itself can in turn be raised to another power, which can typically be chosen as the inverse of said predefined power. If the terms of the sum are, for example, raised to a power of 4, kurtosis is obtained, if the terms are raised to a power of 2, we obtain what is commonly called an RMS or a root mean square.

According to alternative embodiments, the average acoustic amplitude is recalculated for each recorded signal of each successive piece. According to other embodiments, the average acoustic amplitude can be constant, for example can be calculated on a first group of parts initially produced and be considered constant then to simplify the calculations.

At the level of the criterion calculation unit 1 8, the calculation of the criterion can typically be done by comparing in relative value the mathematical signature of the part being manufactured with one, and preferably with, several mathematical signatures of parts previously produced and found to be in conformity. Typically the calculation of the criterion can be done by calculating a succession of differences relative two by two, between the mathematical signature of the current part and one of the mathematical signatures of the preceding parts. Said differences can be taken into relative value, normalized a first time by the current mathematical signature to obtain a first value, and normalized a second time by the mathematical signature of the previous piece to obtain a second value also taken into account in the criteria. Thus, if the criterion takes into account three pieces previously produced, the comparison unit 18 will calculate six values of relative differences between the mathematical signature of the current piece and the mathematical signatures of the three preceding pieces. Each of the relative differences of the mathematical signatures is then compared to a first threshold, and the number of exceedances relative to the threshold is counted. The first comparison threshold may be the same for all the relative differences if they are calculated in absolute value.

The comparisons with respect to the threshold (s) can be made in absolute value. According to other embodiments, it is possible to monitor both the exceeding of an upper and a positive threshold, and the exceeding of a lower and a lower threshold by a lower value. The number of exceedances is then compared to a second threshold "S" by the warning unit 19, to determine whether the part is compliant or not. The second threshold may be for example equal to 1, if no Exceeding relative differences from their first threshold is not accepted.

It is possible to envisage variant embodiments in which, instead of calculating a series of relative differences two by two with respect to the signatures of the preceding parts, a single relative difference, or a pair of relative differences, is calculated from the difference between the mathematical signature of the current piece and an average of mathematical signatures of previous pieces.

It is also possible to envisage variant embodiments, in which the memory 17 stores not mathematical signatures, but each series of acoustic amplitudes of a given number of conformal pieces produced previously.

The criterion calculation unit 1 9 and the statistical processing unit 1 6 can then be grouped, to calculate directly, from the acoustic amplitude values Xj of the signal of the part being shaped, and from signals recorded in the memory 17, a sequence of statistical coherence values each calculated taking into account the series of acoustic amplitude values of the current room and counts the series of acoustic amplitude values of one of the preceding rooms. One can then also compare each of the coherence values with a first threshold, count the number of exceedances with respect to this threshold, and compare this number with a second threshold to decide whether the part is compliant or not.

In addition to calculating a final criterion to determine whether the part is to be considered to be in conformity, the criteria calculation unit 1 8 may provide in graphical form a history of exceedances or not, by higher value, by lower value or by simply in absolute value, relative differences or coherence values calculated for the current part. This graphical information can for example be provided by discretizing the relative values of calculated signature differences. The discretization can be done in two values if one carries out only a comparison in absolute value. In the case where an upper threshold and a lower threshold are used, the discretization can be done with three values, one for a conforming value, one for the overruns, and one for the overruns. At each discretized value, we can assign a graphic display value, for example a symbol, a gray level, a color, and use this graphic value to enter a box in a table that summarizes in the form of successive lines. or successive columns, all the comparison values of the current part with the previous parts selected for comparison.

The operator can thus not only monitor the occurrence or not of an overrun of the overall criterion, but also note whether this overshoot results from a significant difference of the current part with one or more previous parts.

The invention is not limited to the embodiments described and can be declined in many variants.

It is possible to envisage variant embodiments in which a non-conforming mathematical signature is not erased but simply is indexed so as not to be taken into account for the comparisons of the following mathematical signatures.

The number of previous parts taken into account may be different from three, for example may be another number preferably between 1 and 10.

The statistical function retained for calculating the mathematical signature may be another function other than kurtosis or RMS, and different methods of calculating RMS and kurtosis may be used.

Some components of the monitoring unit 7, such as the statistical processing unit 16, and the calculation unit of criterion 1 8, can be implemented in software form on the same computer medium, or on separate computer media, without changing the nature of the invention.

The structure of the electrical, electronic and software means used in the architecture of the monitoring unit 7 has not been detailed, since the means for performing the various functions described (signal acquisition, signal processing, statistical processing, etc. data) exist and are known.

The method can be calibrated to detect risks of damage to the tool, or excessive tool wear, causing instability in the manufacturing process.

The monitoring method may require monitoring of the operator until the tracking memory 17 is populated with a sufficient number of relevant mathematical signatures. However, at each new production start, the process adapts itself to the state of wear, temperature ... of the tool on which the production is launched and to the characteristics of the materials and / or used preforms that may vary from one production batch to another.

The method according to the invention allows a reliable detection of a number of defects related in particular to instabilities of plastic deformation, is inexpensive in calculation, easily calibrated depending on the tool and preforms used, for many tool geometries and of kinematics of formatting.

Claims

claims

1. A method of monitoring the manufacture of a batch of parts stamped by a striking tool (1 0), the method comprising the following steps:

an acquisition of a recorded signal (4), in the form of a series of acoustic amplitudes as a function of time, by one or more piezoelectric sensors (2) placed on the tool, or on a struck matrix by the tool,

a step of temporal analysis of the recorded signal (4), for calculating from the series of acoustic amplitudes (X _n ) composing this recorded signal, at least one mathematical signature (S _p ) in the form of a scalar value associated with said piece (20 _p ), and resulting from a statistical type calculation on the sequence of acoustic amplitudes (X _n ),

a calculation of a criterion value (Crit _p ) of coherence as a function of the mathematical signature of the piece (S _p ) and as a function of at least one mathematical signature (S _{p - 1} , S _{p -2} , S _p-

3, S _{p -4} ) of at least one other piece belonging to an evolutive group of pieces previously shaped on the same tool (1 0),

a comparison of the value (Crit _p ) of the criterion of coherence with at least one threshold (S),

the emission of an alert if the criterion of coherence exceeds the threshold,

for each piece for which a criterion value (Critp) of coherence has been calculated, a step of redefinition of the evolutive group of pieces, as a function of the value of the coherence criterion.

2. The method as claimed in claim 1, in which the calculation of the criterion of coherence (Crit _p ) is carried out taking into account the mathematical signature (S _p ) of the workpiece in progress. manufacturing, as well as a predefined number of mathematical signatures (S _{p -1} , S _{p -2} , S _{p -3} , S _{p -4} ) calculated for parts of the evolutive group of parts previously formatted on the tool.

3. Method according to one of claims 1 or 2, wherein the criterion of consistency (Crit _p ) takes into account a difference between the mathematical signature of the workpiece in process (S _p ) and an average of mathematical signatures d a constant number of pieces among the evolutive group of pieces previously formatted on the tool.

4. Method according to one of claims 1 or 2, wherein the criterion of consistency (Crit _p ) takes into account, for each of the parts of the evolutive group of previously formatted parts, at least one relative difference in values (A (S _p , S _{p -i} )) between the mathematical signature of the piece (S _p ) during manufacture and the mathematical signature of the part of the evolutive group (S _{p -i} ) of pieces.

5. Method according to claim 4, wherein the criterion of coherence (Crit _p ) takes into account a sum, or a sum of absolute values, of the relative differences in value between the mathematical signature of the piece (S _p ) being fabrication and each of the mathematical signatures of the pieces (S _{p -i} ) of the evolutive group of pieces previously formatted on the tool.

6. Method according to claim 4, wherein the coherence criterion takes into account a sum, or a sum of absolute values, of discretized values of the relative differences of value (A (S _p , S _{p -i} )) between the signature. of the piece (S _p ) during manufacture and each of the mathematical signatures (S _{p -i} ) of the pieces of the Evolutionary group of pieces previously formatted on the tool.

7. Method according to any one of the preceding claims, wherein the recorded signal (X _t , X ₂ , X ₃ ... X _n ) is a sequence of time-sampled values X _n , and the mathematical signature:

Where q is the number of successive values X _n of the signal recorded during the formatting of the part being manufactured, k and μ are either constant values or values themselves calculated according to the sequence of values. X _n , a and β are integers or not integers which remain identical for all the shaped pieces.

The method of claim 7, wherein the recorded signal (X _t , X ₂ , X ₃ ... X _n ) is a sequence of time-sampled values, and the mathematical signature of the workpiece being produced is either a root mean square or a kurtosis value of the sequence of values (Xi, X ₂ , X ₃ ... X _n ) forming the recorded signal.

The method according to any one of the preceding claims, wherein the beginning (t ₀ ) of the recording time range ([t ₀ , t _n ]) is calculated from a timing signal representative of the position of a movable portion of the tool (10b) during the forming of the part being manufactured.

10. A device for monitoring the quality of parts shaped by striking in a tool, comprising: at least one acoustic emission sensor (2), configured to acquire a recorded signal (X _t , X ₂ , X ₃ ... X _n ) comprising a series of acoustic amplitudes as a function of time,

a synchronization unit (14) configured to trigger, from a time (t ₀ ) calculated as a function of a signal (6) or a sequence of synchronization signals, the acquisition of a sequence of acoustic amplitudes (X _t , X ₂ , X ₃ ... X _n ) by the sensor (2) for a given duration (Δΐ), to obtain a time-type recorded signal,

a statistical processing unit (1 6) configured to calculate a mathematical signature (S _p ) of the recorded signal, in the form of a scalar value associated with a piece (20 _p ) during manufacture, by a statistical type calculation on the following of acoustic amplitudes (X _t , X ₂ , X ₃ ... X _n ),

a tracking memory (17) configured to store an evolutionary sequence (S _{p -1} , S _{p -2} , S _{p -3} , S _{p -4} ) of one or more mathematical signatures computed previously for an evolutionary group of previously formatted parts on the tool, or to memorize a sliding average of these previously calculated mathematical signatures

A criterion calculation unit (1 8) configured to calculate a coherence criterion value (Crit _p ) according to the mathematical signature (S _p ) associated with the workpiece being produced, and according to at least one mathematical signature (S _{p -1} , S _{p -2} , S _{p -3} , S _{p -4} ) associated with another piece previously produced on the same tool, and

An alert unit (19) configured to compare the consistency criterion (Crit _p ) with at least one threshold (S), and configured to, if the coherence criterion exceeds the threshold (19), trigger an alert, and configured for, if the consistency criterion remains below the threshold, record in the tracking memory (17), the mathematical signature (S _p ) calculated for the part being manufactured, or modify the content of the tracking memory (17) as a function of the mathematical signature (S _p ) calculated for the part being manufactured.