SELF-LEARNING FILTER LIFETIME ESTIMATION METHOD

Abstract

A method for the determination of lifetime of a filter of an electric discharge machine the electrical discharge machine in consideration of a maximum allowable filter pressure, wherein the time measuring unit counts the machining time ts during which an electric discharge machining process is running, a filter pressure sensor measures the filter pressure p(k), preferably with a predetermined sampling interval, the pressure measurement p(k) and the respective sampling time t(k) are stored, and the sampled measurements p(k) and the respective sampling times t(k) are used to determine the parameters of an exponential function which best fits to the plurality of sampled measurements regression analysis. The determined parameters include the filter lifetime tf, which serves to determine the residual time to the filter replacement tr and/or the calendar deadline of filter expiration.

Claims

1. A method for the determination of the lifetime of a filter of an electrical discharge machine, comprising wherein a) a maximum allowable filter pressure pmax is stored, b) a machining time measuring unit (19) counts a machining time ts during which an electrical discharge machining process is running, c) during machining time ts a filter pressure sensor (13) repeatedly measures a filter pressure p(k) at sampling times t(k), with a predetermined sampling interval, d) the filter pressure measurements p(k) and the respective sampling time t(k) are stored, e) the parameters of an exponential function, being the time constant Tau and filter lifetime tf, which best fit to the sampled filter pressure measurements p(k) and the respective sampling times t(k) are determined by regression analysis.

2. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein the exponential function matches with the latest sampling point p(k),t(k) by setting an error function .Math.(k) to zero.

3. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein a residual time to filter replacement tr is computed by subtracting the current machining time ts from the calculated filter lifetime tf.

4. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein a calendar deadline is computed by adding the residual time to filter replacement tr to the current date and time.

5. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein a time window Tw is set to specify a desired identification period of latest pressure measurements p(k), and that the pressure measurements p(k) and the respective sampling times t(k) in said time window Tw are used to determine the parameters of an exponential function.

6. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein after the replacement of the filter the machining time ts is reset by the user or automatically reset by the control unit.

7. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein, once the filter pressure p(k) exceeds a predefined threshold pressure pn, repeatedly comparing the current filter pressure p(k) at sampling time t(k) with at least one earlier filter pressure p(k1) at sampling time t(k1), and if the currently measured filter pressure p(k) is lower than at least one earlier filter pressure p(k1), then the control unit executes one or more of the following actions: releases a warning message for the user to inform that the machining time ts has not been reset, and/or; automatically resets the machining time ts.

8. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein, it the latest earlier measured pressure p(k1) was higher than a predefined threshold pu1, and if the currently measured filter pressure is above of a lower pressure threshold pu2l being lower than an initial value of the filter pressure pa, and if the currently measured filter pressure is below of a upper pressure threshold pu2u being higher than an initial value of the filter pressure pa, then the control unit executes one or more of the following actions: releases a warning message for the user to inform that the machining time ts has not been reset, and/or; automatically resets the machining time ts.

9. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein, if the latest earlier measured pressure p(k1) was higher than a predefined value pbl, and if the currently measured filter pressure is below a predefined value pb2, then the control unit executes one or more of the following actions: releases a warning message for the user signaling a problem with the filtration circuit, and/or; pauses the machining process before starting a new pass, or inhibit the start of the machining process.

10. A method for the determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein after the replacement of the filter (ts=0) the filter lifetime tf is set at the value of a predefined overall filter lifetime tfo, and that the set value for the filter lifetime is maintained until a transition pressure threshold pc is reached, and that thereafter the filter lifetime tf is determined based on stored pressure samplings p(k),t(k).

11. A method for the continuous determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein the filter lifetime tf, and/or the residual time to filter replacement tr, and/or the calendar deadline are displayed on a display of the electrical discharge machine.

12. A method for the continuous determination of the lifetime of a filter of an electrical discharge machine according to claim 1, wherein the control unit determines a machining duration of a machining and/or machining of a current workpiece, and that the residual time to filter replacement is compared with said machining duration, and that, if it is determined that the machining duration exceeds the residual time to filter replacement tr the control unit executes one or more of the following actions: releases a warning message for the user to inform about the residual lifetime and/or the need to replace the filter, and/or; pauses the machining process before starting a new pass, or inhibit the start of the machining process, determine a suitable machining sequence of the main cuts and trim cuts, to complete at least one machining or pass.

13. A method for the determination of the lifetime of a filter of an electrical discharge machine, comprising wherein a) a maximum allowable filter pressure pmax is stored, b) the initial calendar date and time tf_a0 at replacement of the filter is stored, c) a filter pressure sensor (13) repeatedly measures a filter pressure p(k) at sampling times ta(k), with a predetermined sampling interval, d) the filter pressure measurements p(k) and the respective sampling time ta(k) are stored, e) the parameters of an exponential function, being the time constant Tau_a and calendar filter lifetime tf_a which best fits to the sampled filter pressure measurements p(k) and the respective sampling times ta(k) are determined by regression analysis, f) the calendar deadline corresponding to the expected filter expiration date and time is computed by adding the calendar filter lifetime tf_a to the initial calendar date and time tf_a0.

14. An electrical discharge machine comprising a control unit (15) wherein the control unit (15) includes a machining time measuring unit (19) and a memory unit (18), a dielectric unit for the conditioning of a dielectric fluid, wherein the dielectric unit comprising a filtration circuit with one or more filters (12) to filter the debris produced by the electrical discharge machining process, wherein the filtration circuit further comprising a filter pressure sensor (13) located in the filtration circuit through which the machining fluid is supplied to the filter and configured to measure a filter pressure, whereas the machining time measuring unit (19) which counts the machining time ts where the electric discharge machining process is producing debris, a filter pressure sensor (13) wherein the filter pressure sensor is configured to measure the filter pressure during the machining time ts, wherein the control unit (15) is configured to sample the filter pressure by said filter pressure sensor (13), wherein the memory unit (18) is configured to store the current machining time ts determined by the machining time measuring unit (19) and the current filter pressure determined by the filter pressure sensor (13), wherein the control unit (15) is configured to calculate a filter lifetime tf based on the stored machining times ts and filter pressure values and a maximum allowable filter pressure pmax, and a display unit (16) is configured to display the calculated filter lifetime tf.

15. A method of using the determination of the lifetime of a filter of an electrical discharge machine according to claim 13, wherein a) the initial calendar date and time tf_a0 at replacement of the filter is stored, b) the calendar deadline corresponding to the expected filter expiration date and time is computed by adding the calendar filter lifetime tf_a to the initial calendar date and time tf_a0.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The invention will now be further described, by way of examples, and with reference to the accompanying drawings, in which:

[0074] FIG. 1 is a block diagram showing a preferred embodiment of the present invention applied to a Wire EDM machine

[0075] FIG. 2a is a typical graph representing the filter pressure over the time, when pressure is measured before the filters.

[0076] FIG. 2b is a typical graph representing the dielectric flow rate through the filtration circuit over the time.

[0077] FIG. 3 is a typical graph representing the filter pressure as a function of the time, showing two behavioral zones.

[0078] FIG. 4 is a typical graph representing the filter pressure as a function of the time, showing the machining time, the filter lifetime, the residual time to filter replacement and the calendar deadline.

[0079] FIG. 5 is an example of pressure measurement, for which tf and tau are calculated. Then the identified pressure function is drawn, using the values of tf and tau.

DETAILED DESCRIPTION

[0080] An EDM machine needs a dielectric unit providing a sufficient quantity of conditioned dielectric fluid to perform the EDM process. An important function of said dielectric unit is the filtration of the dielectric fluid.

[0081] FIG. 1. represents a block diagram of a typical wire electrical discharge machine (Wire EDM). The Wire EDM machine comprises a dielectric unit with a dirty dielectric tank 10 to collect the used dielectric fluid, a clean dielectric tank 1, to collect the conditioned dielectric fluid, and a filtration circuit. The pump 3 is used to fill the work tank 8 and to maintain a desired dielectric level in said work tank. In this representation pumps 2 and 4 are used to feed the dielectric fluid to the upper and the lower nozzle with high pressure. In practice (not shown) there may be one high pressure pump controlled by a frequency converter and then valves to control the dielectric fluid fed to the upper and the lower nozzle to flush the machining gap. The CNC, the axes control, and the EDM Generator (not shown) are used together to control the inter-electrode distance, that is the distance between the work piece and the wire electrode (both not shown), and to remove material according to a part program stored in the CNC. Due to the material removal occurring during the electrical discharge machining process, the dielectric is continuously charged with debris which are then retained by the filters.

[0082] The filtration circuit has an intake in the dirty dielectric tank 10, from which the dirty dielectric is fed by means of a filtration circuit pump 11 through one or more filters 12, whereby the particles are retained. The filters are cartridges which collect the particles and have a limited lifetime. The filtration circuit ends in the clean dielectric tank 1 where the clean dielectric is stored.

[0083] The filtration circuit further comprises a pressure sensor 13, and/or a flow rate sensor 14 which is arranged in series to the filter 12. The measurement provides information about the actual state of the filter 12, that is, its increasing hydraulic resistance due to the progressive occlusion.

[0084] The control unit 15 comprises or is connected with a machining time measuring unit 19 which counts the machining time ts during which the electrical discharge machining process is running. The control unit 15 stores the measured values in the memory 18 to keep a record of filter pressure values p(k) sampled during the machining process. The actual time t(k) of each sampling in the machining timescale ts will also be recorded in memory 18 by the control unit 15. The stored measurements constitute the history of the filter state. The memory 18 is reset at each replacement of the filters, either by the user or by the control unit. At the reset of the machining time ts=0 an initial value of tf=tfo is set, tfo representing a predefined overall filter lifetime. This value tfo is used to compute the residual time to filter replacement, until a predetermined transition pressure threshold pc is reached. When the filter pressure reaches the threshold pc, the filter lifetime tf is computed using the data stored in memory 18, the control unit 15 computes the filter lifetime tf, the residual time to filter replacement tr, and the calendar deadline of the estimated end of life of the filters 12. The control unit 15 displays these results on the display 16.

[0085] According to a preferred embodiment of the invention, a pressure sensor 13 is placed before the input of the filters 12, that is, at the filter entry side. Here the values measured by the pressure sensor increase progressively with the machining time due to the progressive obstruction of the filter and consequent increase of the hydraulic resistance. FIG. 2a is a qualitative representation of the filter pressure as a function of the time, for the above mentioned embodiment.

[0086] According to another embodiment the pressure drop, respectively the pressure differential of the pressure before and after the filter is measured by means of a differential pressure sensor, so that only the hydraulic resistance of the filter is observed. FIG. 2a applies also to this case.

[0087] According to a further embodiment, a flow rate sensor 14 is placed in the filtration circuit, preferably at the output of the filters 12. FIG. 2b is a qualitative representation of the flow rate across the filter as a function of the time.

[0088] FIG. 3 is a qualitative representation of the filter pressure as a function of the time. Here the pressure sensor is placed at the filter entry side; this is the typical case. The curve has a substantially monotonically increasing progression which is proportional to the progressive occlusion of the filter, however the curve shows two behavioral zones. The zone B corresponds to an exponential behavior, for the pressure as a function of the time. Zone A is a transition zone, where the filters are brand new. Here the filter pressure is nearly constant and a correlation with the soiling based on pressure samplings is not reliable.

[0089] According to a preferred embodiment of the invention, the computation of the filter lifetime and residual time to filter replacement is made by the control unit 15, using different rules for the two zones. In zone B a fitting to an exponential function will determine both, a time constant tau and a filter lifetime tf, as shown in FIG. 3. This filter lifetime tf is defined as the time where the filter pressure reaches the maximum allowable filter pressure pmax (see FIG. 3). It is to be noted that t=0 on filter replacement, and t=tf when pressure reaches pmax (end of life). In zone A, which is a transition zone, the filter pressure is not relevant for the computation of the filter lifetime tf, respectively the residual filter lifetime tr; however the filter pressure is measured also in zone A in view of the computation made in zone B which is based on historical samplings. In zone A tf cannot be computed reliably based on the filter pressure value history, but is computed based for instance on the effective lifetime of the previous filters mounted on the machine (using the predefined overall filter lifetime tfo) and by down-counting the machining hours. The switch from the method applicable to zone A, to the method applicable to zone B is made when the filter pressure reaches a transition pressure threshold pc, for instance 0.5 bar (50 kPa) as shown in FIG. 5.

[0090] FIG. 4 displays a graph representing the filter pressure as a function of the time. As mentioned the curve is an exponential. The initial filter pressure pa in the filter circuit is the filter pressure with new filters. This initial filter pressure pa corresponds essentially to the asymptotic of the exponential function. The measured initial filter pressure is not significantly higher than said asymptotic; within the present invention we will consider the two as equivalent. Here the timescale used in the invention is shown with: the filter lifetime tf, the machining time ts, the residual time to filter replacement tr, and the calendar deadline.

[0091] The following example refers to the case in which the filter pressure measured ahead of the filters in flow direction is used as the significant physical value. As mentioned the filter pressure rises essentially according to an exponential law, as follows:

p(k)=pa+(pmaxpa)*exp((t(k)tf+(k)/tau),

where: [0092] k is a sampling index and applies to the recorded pressure p and time t; [0093] t(k) is the sampling time of a sample k over the machining time, such as t=0 at filter replacement. t(k) is expressed in seconds; [0094] p(k) is the measured filter pressure at sampling time t(k). p(k) is expressed in Pascal; [0095] pa is a constant which corresponds essentially to the asymptotic of the exponential function; said asymptotic is nearly equivalent to the initial filter pressure in the filter circuit, when new filters are inserted. pa is expressed in Pascal; [0096] pmax is the maximal allowed filter pressure. pmax is expressed in Pascal; [0097] tf is an output of the algorithm, representing the filter lifetime, that is the time during which the filter is expected to be serviceable. In other words, if the identification algorithm was perfect, the time tf would be given at the same value, from the filter replacement, until the end of filter life. At this precise time t, the machining time is would be equal to tf. tf is expressed in seconds; [0098] tau is another output of the algorithm. tau is expressed in seconds; [0099] is an error function. (k) is its value at sampling time t(k). The goal is to minimize the least squares value of in a specific range. is expressed in seconds.

[0100] Preferably, tf and tau are recalculated at each new sampling, so that the latest measured pressure value is considered. tf and tau are calculated such that: [0101] (k.sub.2)=0, for k.sub.2 being the index of last acquired point; [0102] the RMS value of , on the specific range [k.sub.1 . . . k.sub.2], being the smallest possible, with k.sub.1 such as ts(k.sub.2)ts(k.sub.1) is nearly equal to Tw, ts being the machining time, and Tw being the considered time window.

[0103] The residual time to filter replacement tr before filter depletion is expressed as: tr=tft

[0104] In the above description, the time t can either represent machining time ts, or calendar time ta. Depending on the two time representation, one can compute tf and tau either in a machining timescale, or in a calendar timescale.

[0105] To determine tf and tau, we use the sampling period from k=k.sub.1 to k=k.sub.2:

[00001]$.Math.p\left(k_1\right)=\mathrm{pa}+\left(p.Math..Math.\max -\mathrm{pa}\right)*\exp \left(\left(t\left(k_1\right)-\mathrm{tf}+\mathrm{.Math.}\left(k_1\right)\right)/\mathrm{tau}\right),.Math.p\left(k_1+1\right)=\mathrm{pa}+\left(p.Math..Math.\max -\mathrm{pa}\right)*\exp \left(\left(t\left(k_1+1\right)-\mathrm{tf}+\mathrm{.Math.}\left(k_1+1\right)\right)/\mathrm{tau}\right),.Math..Math.\mathrm{.Math.}$$p\left(k_2-1\right)=\mathrm{pa}+\left(p.Math..Math.\max -\mathrm{pa}\right)*\exp \left(\left(t\left(k_2-1\right)-\mathrm{tf}+\mathrm{.Math.}\left(k_2-1\right)\right)/\mathrm{tau}\right),.Math..Math.p\left(k_2\right)=\mathrm{pa}+\left(p.Math..Math.\max -\mathrm{pa}\right)*\exp \left(\left(t\left(k_2\right)-\mathrm{tf}+\mathrm{.Math.}\left(k_2\right)\right)/\mathrm{tau}\right),$

[0106] As we want the calculation of p(k.sub.2) to be exact at the filter end of life, one finds (k.sub.2)=0 in this case. Using this criterion, even before the filter end of life, one can find tau as follows:

tau=(t(k.sub.2)tf)/ln((p(k.sub.2)pa)/(pmaxpa)).

[0107] Then the error function becomes for k<k.sub.2:

(k)=(q(k.sub.2)q(k))*tft(k)*q(k.sub.2)+t(k.sub.2)*q(k),

with q (k)=ln((p(k)pa)/(pmaxpa)).

[0108] Minimizing the RMS value of in range k.sub.1 . . . k.sub.21, means to solve d/dtf{[((k)).sup.2]}=0.

[0109] One can solve it and find tf:

tf=num/den,

with

num=[(q(k.sub.2)q(k))*(q(k.sub.2)*t(k)q(k)*t(k.sub.2))], k=k.sub.1 . . . k.sub.21

den=[(q(k.sub.2)q(k)).sup.2], k=k.sub.1 . . . k.sub.21

[0110] For tau computation, remember that:

tau=(t(k.sub.2)tf)/ln((p(k.sub.2)pa)/(pmaxpa)).

[0111] FIG. 5 illustrates a numerical example, with the time window Tw=100 h considered for the determination of the parameters of the exponential function, a sampling interval of 20 h, providing 5 samplings, and an initial filter pressure in the filter circuit of pa=0.3 bar (30 kPa):

The following table shows the filter pressure values which have been measured at a certain machining time ts:
ts [h] 300 320 340 360 380
p [bar] 0.8 0.9 0.9 1.0 1.3

[0112] The parameters of the exponential function are thus determined by using the above formulas:

tf=472.2 h; tau=100.6 h

[0113] The invention uses the pressure p(k) measured at sampling times t(k) to determine the parameters of an exponential function which best fits with the pressure samplings, where said parameter comprise the filter lifetime which will be achieved at a given maximum filter pressure based on current samplings. It goes without saying that, instead of the filter pressure, one could sample another value reflecting the filter status to determine the parameters of an exponential function and lastly to determine the filter lifetime and the residual filter lifetime to filter replacement. This may be for instance a filter flow rate or the level of the dielectric in the clean tank. The measuring methods may also be used together.

[0114] As illustrated in FIG. 5 a transition pressure threshold (a trigger level) is set at 0.5 bar. In this example the filter pressure of 0.5 bar is reached after 200 machining hours. During the first 200 h the filter lifetime is determined by the predefined overall filter lifetime tfo. Thus the first rule, i.e. the down-counting method is used during about of the entire filter lifetime. The second rule, i.e. the determination of the exponential function fitting the sampled pressure measurements is used during about of the entire filter lifetime. The filter pressure is sampled during the entire machining time, i.e. also during the initial phase, to get the required sampling points for the regression analysis.

[0115] As shown in the diagram the pressure samplings of the first 80 h are constant, thus the transition pressure threshold pc of 0.5 bar and the time window Tw of 100 h are appropriately set.

[0116] The invention has been described here above referring to a wire EDM however the same method is applicable to other manufacturing machines and processes such as die sinking EDM, EDM drilling, EDM milling, EDM grinding, etc., but also to other processes in which such cartridge filters are used.

[0117] It goes without saying that the method according the invention can be adopted with an arbitrary number of filter cartridges, whereas the filters can be machine specific or shared with other machines of a workshop.

REFERENCE LIST

[0118] 1 clean dielectric tank [0119] 2,3,4 pumps [0120] 7,9 Upper-/lower nozzle [0121] 8 work tank [0122] 10 dirty dielectric tank [0123] 11 filtration circuit pump [0124] 12 Filter(s) [0125] 13 pressure sensor [0126] 14 flow rate sensor [0127] 15 control unit [0128] 16 display unit [0129] 18 memory [0130] 19 machining time measuring unit