WO2022258240A1

WO2022258240A1 - Method for machining a workpiece

Info

Publication number: WO2022258240A1
Application number: PCT/EP2022/058477
Authority: WO
Inventors: Thomas Wuhrmann; Patrick Waibel; Nikola PASCHER; Alwin HAENSEL; Gunnar KEITZEL
Original assignee: Kistler Holding Ag
Priority date: 2021-06-09
Filing date: 2022-03-30
Publication date: 2022-12-15
Also published as: EP4352578A1

Abstract

The invention relates to a method for machining a workpiece (0, 0') into a piece good (1, 1') in multiple machining processes (Bj) using a machine tool (10) with multiple tools (12j). In each machining process (Bj), the tool force (F) exerted onto the workpiece (0, 0') is measured as a chronological sequence of measurement signals (Sj) by means of a piezo electric force sensor (14), wherein in a learning phase (V1, V2, V3, V4), a first workpiece (0) is machined, a quality feature (Q) is checked, and an accepted-part characteristic (GKij) is ascertained and is classified into an accepted-part class (GCj); and in a subsequent phase (V5, V6, V7), another workpiece (0') is machined, and a piece good characteristic (SKij) is ascertained and is classified into the accepted-part class (GCj). Thus, a first workpiece (0) is identified as an accepted part and is classified into an accepted-part class, and for a subsequently machined workpiece (0'), the workpiece characteristic (SKij) is used to ascertain whether the subsequently machined workpiece is classified into the accepted-part class (GCj) without needing to check whether the additional piece good has the quality feature (Q). The classification of the piece good characteristic (SKij) into the accepted-part class (GCj) is used to automatically arrive at two conclusions as to whether the additional piece good is an accepted part or a rejected part and whether required tools are worn or not, and if so which of the required tools are worn.

Description

Process for machining a workpiece

technical field

The invention relates to a method for machining a workpiece. A machine tool, which has a tool, is used for machining the workpiece. The tool exerts a tool force on the workpiece and removes material from the workpiece. In this way, the workpiece is processed into a unit load.

State of the art

A well-known machine tool is a lathe, a milling machine, a sawing machine, etc. The workpiece is made of any material such as metal, wood, plastic, etc. The tool is made of hard, solid and tough cutting material such as metal, ceramics, etc. Well-known tools for machining a workpiece are chisels, turning tools, milling cutters, countersinks, saw blades, etc.

[0003] Now the tool suffers wear and tear when the workpiece is machined. The wear comes from the tool force as well as from increased temperatures during the machining of the workpiece. The wear changes the cutting edge geometry of the tool. A worn tool requires a greater tool force for the machining of the workpiece, which is reflected in a higher energy consumption of the machine tool expressed and which also reduces the surface quality and the dimensional accuracy of the workpiece.

For a cost-effective and qualitatively constant machining of the workpiece, it is therefore desirable to measure the tool force in order to detect wear and tear of the tool at an early stage and to avoid it.

[0005] Document US2019/0210176A1 discloses a method for machining a workpiece, where a large number of parameters of the machine tool are measured several times in a learning phase before the actual machining of the tool. Each set of parameters includes machine tool vibration and noise, tool force exerted by the tool, and machine tool load and power consumption. Each set of parameters is represented as a vector and classified in a vector space using machine learning. Only normal vectors that correspond to an unworn tool are classified in the vector space. The actual machining of the tool follows the learning phase. Here, too, a large number of parameters are measured each time a tool is machined and it is checked whether the large number of measured parameters represent a normal vector. If so, the normal vector is binned into vector space. If this is not the case, the machining of the workpiece is interrupted and the tool is checked for wear. [0006] The object of the present invention is to simplify the method disclosed in US2019/0210176A1 for machining a workpiece.

Presentation of the invention

[0007] This object is achieved by the features of the independent claim.

The invention relates to a method for the machining of a workpiece, which workpiece is processed in a time sequence of several machining operations by several tools required to form a unit load, with a tool force being applied to the workpiece in each machining operation by a required tool exercised, which tool force is measured in each machining process as a time sequence of measurement signals; wherein in a first step of the method a first workpiece is machined to form a first piece good; in a second step of the method, it is checked whether the first piece good has at least one quality feature and is a good part; in a third step of the method, at least one good-part characteristic value is determined for each temporal sequence of measurement signals of the machining operations that led to the good part; wherein in a fourth step of the method for each machining process that has led to the good part, the good part characteristic value is classified into at least one good part class; wherein in a fifth step of the method, which takes place chronologically after the first four steps of the method, a further workpiece is processed to form a further piece good; wherein in a sixth step of the process at least one piece goods characteristic value is automatically determined for each time sequence of measurement signals of the processing operations that led to the further piece goods; and wherein in a seventh step of the method for each processing operation that has led to the further piece goods, the piece goods characteristic value is automatically classified in the good part class.

The inventors have surprisingly found out that to determine whether a tool is worn out during machining of a workpiece or not, a large number of parameters of the machine tool do not have to be measured, but that it is sufficient here for the To measure tool force, which simplifies the method according to the invention in comparison with that of the document US2019/0210176A1.

However, the difficulty to be overcome has been that the tool force in the machining operations due to the different types of machining such as roughing, drilling, turning, finishing, parting has various large amounts and the amounts of the tool force also vary greatly. Only by reducing the time sequence of measurement signals to a good part characteristic and by classifying the good part characteristic in a good part class is the tool force for machining processes with different types of machining qualitatively comparable.

[0011] In a learning phase, a first workpiece that has been turned into a good part in several machining processes leads has determined a good part parameter. And this good part parameter is classified in a good part class. If another workpiece is now processed, only a piece goods characteristic value has to be determined for this further piece goods characteristic value is then automatically classified into the good part class without having to check whether the further piece goods have at least one quality feature , because if the piece goods parameter cannot be classified into the good part class, the part is a bad part.

[0012] With the classification of the piece goods characteristic value in the good part class, two statements are automatically made, that the other piece goods are a good part or a bad part, and whether and which of the required tools is closed or not.

Further developments of the subject invention are claimed in the dependent claims.

Brief description of the figures

The invention is explained in more detail below by way of example with reference to the figures. Show it

1 shows a flowchart with steps VI to V7 of the method for machining a workpiece 0, 0' into a piece good 1, 1';

Fig. 2 shows a schematic view of parts of a machine tool 10, a profilometer 20 and an evaluation unit 30 for carrying out the method according to FIG. 1; and Fig. 3 a decision tree Ej,

for carrying out steps V4 and V7 of the method according to FIG.

The same reference symbols designate the same objects in the figures.

Ways to carry out the invention

1 shows a flow chart with several steps VI-V7 of the method for machining a workpiece 0, 0'. The workpiece 0, 0' is finished in a chronological sequence of several machining operations Bj to form a unit load 1, 1'. The editing process-

Index j is a positive integer m, j=l...m. The machining of the workpiece 0, 0' takes place with several required tools 12j, j=l...m. In every machining process

Bj, j=l...m, a tool force F is exerted on the workpiece 0, 0' by a required tool 12j, j=l...m. the

Tool force F is measured in each machining process Bj, j=l...m as a time sequence of measurement signals Sj, j=l...m.

In a first step VI of the method, a first workpiece 0 is machined into a first item 1 .

In a second step V2 of the method, it is checked whether the first piece good 1 has at least one quality feature Q and whether it is a good part G.

In a third step V3 of the method, for each chronological sequence of measurement signals Sj, j=l...m of the machining processes Bj, j=l...m that led to the good part G ben, at least one good part parameter GKij, i=l...n,

determined.

In a fourth step V4 of the method, for each processing operation Bj, j=l...m, the good-part characteristic value GKij, i=l...n, j=l...m is put into a good-part class GCj , j=l...m classified.

In a fifth step V5 of the method, which takes place after the first four steps VI-V4 of the method, a further workpiece 0' is processed into a further piece goods 1'.

In a sixth step V6 of the method, for each time sequence of measurement signals Sj, j=l...m of the processing operations Bj, j=l...m that have led to the further piece goods 1', at least one Piece goods parameter SKij, i=l...n, j=l...m automatically determined.

In a seventh step V7 of the method, the piece goods characteristic value SKij, i=l...n, j=l...m is automatically assigned to the good part class for each processing operation Bj, j=l...m GCj, j=l...m is classified.

Fig. 2 shows a machine tool 10 and an evaluation unit 20 for carrying out the method for machining the workpiece 0, 0' into an item 1, 1' according to Fig. 1.

The workpiece 0, 0 'is made of any material such as metal, wood, plastic, etc. The workpiece 0, 0' summarizes at least one first workpiece 0 and at least one another workpiece 10'. The piece goods 1, 1' comprise at least one first piece good 1 and at least one further piece good

1 .

The machine tool 10 is arranged in an orthogonal coordinate system with three axes x, y, z, the three axes are also referred to as the transverse axis x, horizontal axis y and vertical axis z. The machine tool 10 has a tool holder 11, several required tools 12j, j=l...m and a tool carriage 13. The tool holder 11 holds the required tools 12j, j=l...m at the same time in different machining positions p=l...m. The machining position p is a positive integer m (p=l...m). The tool slide 13 moves the tool holder tion 11 along the three axes x, y and z. In this way, the process for machining the workpiece 0, 0' into a finished unit good 1, 1' does not have to be interrupted in order to change a tool, since all the required tools 12j, j=l...m are removed from the tool holder 11 is already worn and by moving the tool holder 11 through the tool carriage 13, one of the required tools 12j, j=l...m is brought into contact with the workpiece 0, 0'.

The required tools 12j, j=l...m are used to machine the workpiece 0, 0' in a chronological sequence of several machining processes Bj, j=l...m. The required tools 12j, j=l...m carry out various types of machining such as roughing, drilling, turning, finishing, parting off, and so on. For example, a first machining operation Bj, j=l consists of roughing, a second machining operation Bj, j=2 from drilling, a third machining operation Bj, j=3 from turning, a fourth machining operation Bj, j=4 from finishing, a fifth machining operation Bj, j=5 from parting off.

The required tools 12j, j=

perform a cutting movement in a transverse plane spanned by the transverse axis x and the vertical axis z during the machining of the workpiece 0, 0'. The required tools 12j, j=l...m also carry out a feed motion along the horizontal axis y during the machining of the workpiece 0, 0'. The required tools 12, j=l...m exert a tool force F on the workpiece 0, 0'. The tool force F has three force components. Due to the cutting movement, the tool force F has a main cutting force component Fx in the direction of the transverse axis x and a passive force component Fz in the direction of the vertical axis z. And due to the feed movement, the tool force F has a feed force component Fy in the direction of the horizontal axis y.

The tool force F has different magnitudes for each of the different types of machining such as roughing, drilling, turning, finishing, parting off, and the magnitudes also fluctuate to different extents. In the sense of the invention, the adjective "qualitative" refers to the quality of the finished piece goods 1, 1', whether the piece goods 1, 1' is a good part or not, and to the quality of each of the required tools 12j, j= l...m whether one of the required tools 12j, j=l...m is closed or not. The machine tool 10 has at least one force transducer 14 . The force transducer 14 is spatially arranged between the tool holder 11 and the tool carriage 13 . The force transducer 14 is held by the tool holder 11. The force transducer 14 measures the tool force F.

The force transducer 14 is preferably a piezoelectric force transducer.

The piezoelectric force transducer comprises piezoelectric material made of single crystal such as quartz (S1O2), calcium gallo-germanate (Ca ₃ Ga2Ge40i4 or CGG), langasite (La3Ga5SiOi4 or LGS), tourmaline, gallium orthophosphate, etc. and piezoelectric ceramics such as lead -zirconate-titanate (Pb [Zr _x Tii- _x ]0 ₃ , 0<x<l), etc. The tool force F to be measured acts on the surfaces of the piezoelectric material. Under the action of the tool force F, the piezoelectric material generates piezoelectric charges. The amount of piezoelectric charges is proportional to the size of the tool force F. The sensitivity of the piezoelectric material is typically 7pC/NDh for a tool force F of 150N, 1050pC electric polarization charges are generated. The amount of piezoelectric charges per measurement is the measurement signal.

The piezoelectric force transducer measures the tool force F with a temporal resolution of less than/equal to 0.01Hz. This means that the piezoelectric force transducer measures the tool force F at least 100 times per time unit of lsec and supplies at least a number z=100 measurements. For a work process Bj,

For a period of 20 seconds, the piezoelectric force transducer measures the tool force F at least 2000 times and supplies at least a number z=2000 measurements. In each processing operation Bj, j=l...m, a time sequence of measurement signals Sj, j=l...m is measured.

The piezoelectric force transducer has electrodes. The electrodes are made of electrically conductive material such as copper, gold, etc. Each electrode is arranged directly on a surface of the piezoelectric material. The electrical polarization charges are picked up by the electrodes as time sequences of measurement signals Sj, j=l...m.

The piezoelectric force transducer has a measurement signal cable 141 . The electrodes are electrically connected to the measurement signal cable 141 . The tapped time sequences of measurement signals Sj, j=l . . . m are derived via the measurement signal cable 141 .

The force transducer 14 can be a multi-component force transducer that measures the main cutting force component Fx, feed force component Fy and the passive force component Fz. Since the three force components of the tool force F in the machining processes Bj, j=l...m are different depending on the required tools 12j, j=l...m, measurement of all three force components of the tool force F forms the basis for a comparability of the measurement signals Sj, j=l...m of the various machining processes Bj, j=l...m. However, the force transducer 14 can also be a one-component force transducer which only measures the main cutting force component. component Fx. A single-component force transducer is cheaper to buy than a multi-component force transducer. The force transducer 14 can also consist of several force transducers and measure the tool force F redundantly. A redundant force measurement has a higher measurement accuracy. In this way, two single-component force transducers can measure the same tool force F. The measurement result is then the arithmetic mean of the temporal sequences of measurement signals Sj, j=l...m of the two single-component force transducers, corrected by a factor of the distance of each of the single-component force transducers from the machining position p, p=l... .m of the required tool 12j, j=l...m, whose tool force F is measured.

The profilometer 20 is a measuring device for three-dimensional measurement of a surface topography. The profiler 20 is a diamond probe, an interferometer, etc. With the profiler 20 at least one quality feature Q of the first piece 1 is measured. The quality feature Q can be a surface quality of the first piece good 1, a dimensional accuracy of the first piece good 1, and so on. In the example according to FIG. 2, the profilometer 20 is an interferometer that optically measures the surface topography of the first piece good 1 with a measurement accuracy in the range of +/-1pm. The profiler 20 generates quality signals QS for the measured quality feature Q. The quality signals QS are preferably digital data. The profilometer 20 has a quality signal cable 201 . The quality signals QS are derived via the quality signal cable 201. The evaluation unit 30 has at least one computer 301, at least one converter unit 302, at least one interface 303, at least one input unit 304 and at least one output unit 305.

The computer 301 has at least one data processor and at least one data memory. At least one computer program C is stored in the data memory and can be loaded into the data processor. The computer program C loaded into the data processor causes the computer 301 to carry out a step of the method. The computer 301 performs a step of the method automatically. For the purposes of the invention, the adjective "automatically" means carrying out a step of the method without the involvement of a person. For this purpose, the computer program C generates commands.

The force transducer 14 transmits the time sequences of measurement signals Sj, j=l...m to the converter unit 302. The force transducer 14 is preferably electrically connected to the converter unit 302 via the measurement signal cable 141. The converter unit 302 converts the time sequences of measurement signals Sj, j=l...m into time sequences of measured values Xj, j=l...m. The converter unit 302 is preferably an electrical charge amplifier which automatically converts temporal sequences of measurement signals Sj, j=l...m in the form of electrical polarization charges into electrical direct voltages and automatically digitizes them. The time series of measured values Xj, j=l...m are digital data. The chronological sequences of measured values Xj, j=l...m can be loaded into the data processor. The temporal sequences of measured values Xj, j=l...m can be stored in the data memory. The profilometer 20 is connected to the interface 303 via the quality signal cable 201 . The quality signals QS can be loaded into the data processor from the interface 303 . The quality signals QS can be stored in the data memory by the interface 303 .

The computer 301 can be operated via the input unit 304. The input unit 304 can be a keyboard for entering commands. For the purposes of the invention, the verb "operate" means that the computer 301 is started, controlled and switched off by a person using the input unit 304 with commands. The person is a person. The person is trained to use the machine tool 10, the profilometer 20 and to operate the evaluation unit 30. The output unit 305 can be a screen for graphically displaying digital data.

The second step V2 of the method can be carried out by the person. For this purpose, at least one quality feature of a good part G is stored as quality data QD in the data memory. The person commands the computer 301 to load the quality signals QS into the data processor. At least one quality feature of a good part G is stored as quality data QD in the data memory. The quality data QD is digital data. The person commands the computer 301 to load the quality data QD into the data processor. The person commands the computer 301 to compare the quality signals QS with the quality data QD.

The second step V2 can also be carried out automatically by the computer 301. This causes the in the computer program C loaded into the data processor directs the computer 301 to carry out the second step V2 of the method. The computer program C commands the computer 301 to load the quality signals QS into the data processor. The computer program C commands the computer 301 to load the quality data QD into the data processor. The computer program C commands the computer 301 to compare the quality signals QS with the quality data QD.

The comparison can have a positive result or a negative result. If the result is positive, the quality signals QS correspond to the quality data QD within the scope of the measuring accuracy of the profilometer 20 . If the result is negative, the quality signals QS deviate from the quality data QD within the scope of the measuring accuracy of the profilometer 20 . The comparison is therefore a check as to whether the first piece good 1 has at least one quality feature Q and is a good part G. The computer 301 graphically displays the positive result or the negative result on the output unit 305. The computer 201 stores the positive result or the negative result as digital data in the data memory.

The computer program C loaded into the data processor causes the computer 301 to carry out the third step V3 of the method. The computer program C orders the computer 301 for each temporal sequence of measurement signals Sj, j=l...m of the processing operations Bj, j=l...m that led to the good part G, at least one good part characteristic value GKij, i =l...n, j=l...m to be determined. For this purpose, the computer program C commands the computer 301 for each processing operation Bj, j=l...m which has led to a good part G to load the chronological sequence of measured values Xj, j=l...m into the data processor .

The computer program C now orders the computer 301 for each time sequence of measured values Xj, j=l...m at least one of the following good part characteristic values GKij, i=l...n, j=l... m to determine. The characteristic index i is a positive integer i=l...n, n is the characteristic number.

A first good-part characteristic value GKI is a good-part mean value GKlj, j=l...m of the chronological sequence of measured values Xj, j=l...m of a machining process Bj, j=l...m.

A second good-part characteristic value GK2 is a smallest good-part value GK2j, j=l...m of the chronological sequence of measured values Xj, j=l...m of a machining operation Bj, j=l...m .

A third good part characteristic value GK3j, j=l...m is a largest good part value GK3 of the chronological sequence of measured values Xj, j=l...m of a machining operation Bj, j=l...m.

A fourth good-part characteristic value GK4j, j=l...m is a good-part measurement duration value GK4 of the chronological sequence of measured values Xj, j=l...m of a machining process Bj, j=l...m.

A fifth good part characteristic value GK5j, j=l...m is a good part measurement time deviation value GK5 of the chronological sequence of measured values Xj, j=l...m of a machining process Bj, j=l...m with respect to a Standard measurement duration of a machining process Bj, j=l...m. A sixth good part characteristic value GK6j,

is a

Good part measured value deviation value GK6 of the time sequence from

of a machining process Bj,

ines standard measured value of a machining process Bj,

A seventh good part parameter GK7j,

is a

Good part curve integral value GK7 of the time sequence of measured values Xj,

of a machining operation Bj,

An eighth good part parameter GK8j,

is a

Good part expected value GK8 of the time sequence of measured values Xj, j=

of a machining operation Bj,

regarding the standard measured value of a machining operation Bj, j=l...m. The chronological sequence of good part expected values GK8 is standardized to zero.

Thus, for each machining process Bj, j=l...m that has led to the good part G, the chronological sequence of measured values Xj, j=l...m is based on a good part characteristic value GKij, i= l...n, j=l...m reduced.

The computer program C orders the computer 301 for each processing operation Bj, j=1...6 the good part characteristic value GKij, i=l...n, j=l...m in the data memory as the existing good part to store characteristic value GKij, i=l...n, j=l...m.

The computer program C loaded into the data processor causes the computer 301 to carry out the fourth step V4 of the method. The computer program C commands the computer 301 the at least one good part characteristic value GKij, i=l...n,

to classify a good part class GCj, i=l...m.

The good part class GCj, j=l...m is determined by machine learning. Preferably, a monitored machine learning takes place with a decision tree Ej, j=l . . . m for the graphical representation of hierarchically consecutive decisions. Fig. 3 shows an example of a decision tree Ej, j=l...m. With the decision tree Ej, j=l...m, the first unit good 1 is classified as a good part G by the good part parameter GKij, i=l...n, j=l...m. A good part class GCj, j=l...m is sought. The adjective "monitored" means that the data necessary for the decision, such as a good part characteristic value GKij, i=l...n, j=l...m, attributes, decision rules and the good part class GCj, j= l...m are provided on the computer and stored in the data memory.The attributes and decision rules are formulated expert knowledge and explain why the first general good 1 is replaced by the good part characteristic value GKij, i=l...n, j=l... .m corresponds to a good part and is therefore classified in the good part class GCj, j=l...m Each decision tree Ej, j=l...m has a root node Elj, j=l...m, paths E2j , j=l...m, interior nodes E3j, j=l...m and leaves E4j, j=l...m Preferably, the decision tree Bj, j=l...m is a binary tree with two paths E2j, j=l...m leading from the root node Elj, j=l...m to two interior nodes E3j, j=l...m From each interior node E3j, j=l. ..m two paths E2j, j=l...m lead to at least one further inner node E3j, j=l...m or to at least one Bl att E4j, j=l...m. The computer program C commands the computer 301 for each processing operation Bj, j=l...m that has led to a good part G, the at least one good part characteristic value GKij, i=l...n, j=l ...m to load into the data processor. The computer inputs the good part index GKij, i=l...n, j=l...m into the root node Elj, j=l...m. The computer program C commands the computer 301 to enter attributes stored in the data memory into the interior nodes E3j, j=l...m. The attributes are assigned to the good part parameter GKij, i=l...n, j=l...m. The attributes identify the good part characteristic value GKij, i=l...n, j=l...m as a good part G or as a bad part S. The computer program C commands the computer C a decision rule stored in the data memory for the good part characteristic value GKij, i=l...n, j=l...m and the associated attributes apply. This application of the decision rule leads to a decision E with two answers. Each response can be another interior node E3j, j=l...m or a leaf E4j, j=l...m. If the answer is another interior node E3j, j=l...m, the computer program C commands the computer 301 to load another attribute stored in data memory into the interior node E3j, j=l...m and the decision tree Bj, j=l...m recursively. If the answer is a leaf E4j, j=l...m, the required good part class GCj, j=l...m is available.

Preferably a whole forest with many decision trees Ej , j=l...m is used. Through regression, a augmented decision tree Ej,j=l...m learns from previous decision trees Ej,j=l...m. The earlier decision trees Ej, j=l...m are weak decision trees that did not lead to the good part class GCj, j=l...m, or they are strong decision trees that have led to several good part classes GCj, j=l...m. Only a classification into a single good part class GCj, j=l...m is also a very accurate classification. The computer program C commands the computer 301 to form a reinforced decision tree Ej, j=l...m by means of a targeted selection and combination of the strong decision trees. The regressive formation of reinforced decision trees Ej, j=l...m is continued until a single good part class GCj, j=l...m is available as a leaf.

[0063]Preferably, several good-part characteristic values GKij, i=l...n, j=l...m are classified into the good-part class GCj, j=l...m. This improves the quality of the classification. For the purposes of the invention, the term "quality" denotes the probability with which a good part G is actually classified in the good part class GCj, j=l...m.

The computer program C commands the computer 301 for each processing operation Bj, j=1...6 the good part class GCj, j=l...m in the data memory as an existing good part class GCj, j=l... m to save.

The learning phase of the first four steps VI-V4 of the method is preferably repeated several times. For each processing operation Bj, j=l...m, an arithmetic mean is formed for the at least one good part characteristic value GKij, i=l...n, j=l...m, which is repeated several times. According to the law of large numbers, the good part characteristic value GKij, i=l...n, j=l...m, which is repeated several times, converges to the arithmetic mean, which yields the most probable good part characteristic value GKij, i=l... n, j=l...m. The computer program C commands the computer ter 301 for each processing operation Bj, j=1 to be saved as an existing good part parameter GK+i, i=l...n. The computer program C orders the computer 301 an optimized good part class GC+i, i for each processing operation (Bj, j=l...m) for the at least one most probable good part characteristic value GK+i, i=l...n =l...n to determine. The computer program C commands the computer 301 to store the optimized good part class GC+i, i=l...n in the data memory. The optimized good part class GC+i, j=l...n is used as the good part class GCj, j=l...m in the seventh step V7 of the method.

The optimized good part class GC+i, i=l...n has the highest sensitivity when classifying piece goods 1, 1′ as good part G or bad part S. As a result, the optimized good part class GC+i, i=l...n reduces the probability that a piece good 1', 1' will be classified as false positive or false negative. This keeps the manufacturing costs as low as possible, particularly for piece goods 1, 1' that are expensive to manufacture.

For a workpiece 0' machined in the fifth step V5 of the method to form a further piece good 1', in the sixth step of the method, in the sixth step V6 of the method for each time sequence of measurement signals Sj, j=l...m of the processing operations Bj, j=l...m which have led to the further piece goods 1', at least one piece goods characteristic value SKij, i=l...n, j=l...m is automatically determined. For this purpose, the computer program C commands the computer 301 to load the time sequence of measured values Xj, j=l...m into the data processor for each processing operation Bj, j=l...m that has led to the item 1 .

The computer program C now commands the computer 301 for each time sequence of measured values Xj, j=l...m at least one of the following piece goods characteristic values SKij, i=l...n, j=l... m to determine. The characteristic index i is a positive integer i=l...n.

A first piece goods characteristic value SKlj, j=l...m is a piece goods mean value SKI of the chronological sequence of measured values Xj, j=l...m of a processing operation Bj, j=l...m.

A second piece goods characteristic value SK2j, j=l...m is a smallest piece goods value SK2 of the chronological sequence of measured values Xj, j=l...m of a processing operation Bj, j=l...m .

A third piece goods characteristic value SK3j, j=l...m is a largest piece goods value SK3 of the chronological sequence of measured values Xj, j=l...m of a processing operation Bj, j=l...m .

A fourth piece goods characteristic value SK4j, j=l...m is a piece goods measurement duration value SK4 of the chronological sequence of measured values Xj, j=l...m of a processing operation Bj, j=l...m.

A fifth piece goods characteristic value SK5j, j=l...m is a piece goods measurement time deviation value SK5 of the chronological sequence of measured values Xj, j=l...m of a processing operation Bj, j=l...m a standard measuring time duration of a machining process Bj, j=l...m. A sixth piece goods parameter SK6j,

is a piece goods measured value deviation value SK6 of the time sequence of measured values Xj,

of a machining operation Bj, with respect to a standard measurement value of a machining operation Bj,

A seventh piece goods parameter SK7j,

is a

Piece goods curve integral value SK7 of the time sequence of measured values Xj,

of a machining operation Bj, j=1...m.

An eight piece goods parameter SK8j,

is a

Expected piece goods value SK8 of the time sequence of measured values Xj, j=

of a machining operation Bj,

regarding the standard measured value of a machining operation Bj, j=l...m. The chronological sequence of expected piece goods values SK8 is standardized to zero.

The computer program C commands the computer 301 for each processing operation Bj, j=1...6 that has led to the piece good 1, the piece good characteristic value SKij, i=l...n, j=l...m in the

store the data memory as an existing piece goods parameter SKij, i=l...n, j=l...m.

The computer program C loaded into the data processor causes the computer 301 to carry out the seventh step V7 of the method.

For this purpose, the computer program C commands the computer 301 for each processing operation Bj, j=l...m which has led to the further piece good 1, the piece good characteristic value SKij, i=l...n,

and load the existing good part class GCj, j=l...m into the data processor.

The computer program C commands the computer 301 to classify the piece goods parameter SKij, i=l...n, j=l...m into the good part class GCj, i=l...m. A bad part class is present if the piece goods parameter SKij, i=l...n, j=l...m cannot be classified into the good part class GCj, i=l...m. A decision tree Ej, j=l...m as shown in FIG. 3 with a root node Elj, j=l...m, paths E2j, j=l...m, inner nodes E3j, j=l...m and leaves E4j, j=l...m used.

The computer inputs the piece-part identifier SKij, i=l...n, j=l...m into the root node Elj, j=l...m. The computer program C commands the computer 301 to enter attributes stored in the data memory into the interior nodes E3j, j=l...m. The attributes are assigned to a good part characteristic value GKij, i=l...n, j=l...m, whose characteristic value index i is identical to the characteristic value index i of the piece part characteristic value SKij, i=l... .n, j=l...m. The attributes identify the good part characteristic value GKij, i=l , i=l...n, j=l...m and the associated attributes. This application of the decision rule leads to a decision E with two answers. Each response can be another interior node E3j, j=l...m or a leaf E4j, j=l...m. If the answer is another inner node E3j, j=l...m, the computer program C orders the computer 301 another one in the data memory. stored attribute in the interior nodes E3j,

to load and the decision tree Ej,

continue recursively. If the answer is a sheet E4j,

is, the good part class is GCj,

before. If a piece goods characteristic value SKij, i=l...n, j=l...m cannot be classified into the good part class GCj, j=l...m, there is a bad part.

reference list

0, 0' workpiece 1, 1 piece goods

10 machine tool

11 tool holder

12j required tool

13 tool slides

14 force transducer 141 measuring signal cable 20 profilometer 201 quality signal cable 30 evaluation unit

301 computers

302 converter unit

303 interface

304 input unit

305 Output unit Bj Machining process C Computer program

E decision Ej decision tree Elj root node E2j path E3j inner node E4j leaf F tool force

Fx main cutting force component

Fy feed force component

Fz passive power component

G good part

GCij good part class

GC+ij optimized good part class GKij Good part characteristic value GKlj Good part1 average value GK2j Smallest good part value GK3j Largest good part value GK4j Good part1 measurement time duration value GK5j Good part1 measurement time duration deviation value GK6j Good part1 measured value deviation value GK7j Good part curve integral value GK8j Good part expected value GK+ij most probable good part characteristic value

Parameter index j Machining process index

VI - V7 step

N Newton m number n index number

P processing position

Q quality feature

QD quality data

QS quality signal

Sj measurement signal

SKij Piece goods i1 characteristic value SKlj Piece goods mean value SK2j Smallest piece goods value SK3j Largest piece goods value SK4j Piece goods measurement time duration value SK5j Piece goods measurement time duration deviation value SK6j Piece goods measured value deviation value SK7j Piece goods curve integral value SK8j Piece goods expected value xj Measured value z number

Claims

patent claims

1. Method for machining a workpiece (0, 0'), which workpiece (0, 0') in a time sequence of several machining operations (Bj, j = l ... m) of several required tools (12j, j =l...m) is finished to a piece well (1, 1'), thereby in each

Machining operation (Bj, j=l...m) exerted by a required tool (12j, j=l...m) a tool force (F) on the workpiece (0, 0'), which tool force (F) in each Machining process (Bj, j=l...m) as a temporal sequence of measurement signals (Sj, j=l...m) is measured; wherein in a first step (VI) of the method, a first workpiece (0) is processed to form a first piece good (1); wherein in a second step (V2) of the method it is checked whether the first piece good (1) has at least one quality feature (Q) and is a good part (G); wherein in a third step (V3) of the method for each de time sequence of measurement signals (Sj, j = l ... m) of the Be processing operations (Bj, j = l ... m) led to the good part (G). have, at least one good part parameter (GKij, i=l...n, j=l...m) is determined; in a fourth step (V4) of the method for each machining operation (Bj, j=l...m) which led to the good part (G), the good part characteristic value (GKij, i=l...n, j=l...m) is classified into a good part class (GCj, j=l...m); wherein in a fifth step (V5) of the method, which takes place after the first four steps (VI - V4) of the method driving takes place, another workpiece (0 ') to a white direct piece goods (1') is processed; in a sixth step (V6) of the method for each temporal sequence of measurement signals (Sj, j=l...m) of the processing operations (Bj, j=l...m) leading to the further piece goods (1'). have, at least one piece goods characteristic value (SKij, i=l...n, j=l...m) is automatically determined; and wherein in a seventh step (V7) of the method for each processing operation (Bj, j=l...m) that led to the further piece goods (1'), the piece goods characteristic value (SKij, i=l... .n, j=l...m) is automatically classified into the good part class (GCj, j=l...m).

2. Method according to claim 1, characterized in that the first four steps (VI - V4) of the method are repeated several times; that for each processing operation (Bj, j=l...m) for the at least one good part characteristic value (GKij, i=l...n, j=l...m) repeated several times, a most probable good part characteristic value ( GK+i, i=l...n); that for each machining process (Bj, j=l...m) for the most probable good part characteristic value (GK+i, i=l...n) an optimized good part class (GC+i, i=l... n) is determined; and that the optimized good part class (GC+i, j=l...n) is used as the good part class (GCj, j=l...m) in the seventh step (V7) of the method.

3. Method according to one of claims 1 or 2, characterized in that in the fourth step (V4) of the method several good part characteristic values (GKij, i=l...n, j=l...m) be classified into the good part class (GCj, j=l...m); and that in the seventh step (V7) of the method, several piece good parameters (SKij, i=l...n, j=l...m) automatically in the

Good part class (GCj, j=l...m) are classified.

4. Method according to one of claims 1 or 2, characterized in that in the fourth step (V4) of the method, more than five good part characteristic values (GKij, i=l...n, j=l... m) classified into the good part class (GCj, j=l...m); and that in the seventh step (V7) of the method, more than five piece goods characteristic values (SKij, i=l...n, j=l...m) are automatically assigned to the good part class (GCj, j=l... m) be classified.

5. Method according to one of claims 1 to 4, characterized in that in the third step (V3) of the method at least one of the following good part parameters (GKij, i=l...n, j=l...m) is determined: a mean value for a good part (GKij, j=l...m), a smallest good part value (GK2j, j=l...m), a largest good part value (GK3j, j=l... m), a good part measurement duration value (GK4j, j=l...m), a good part measurement duration deviation value (GK5j, j=l...m), a good part measurement deviation value (GK6j, j=l...m) , a good part curve integral value (GK7j, j=l...m), a good part expected value (GK8j, j=l...m).

6. Method according to one of claim 5, characterized in that in the sixth step (V6) of the method at least one of the following piece goods characteristic values (SKij, i=l...n, j=l...m) automatically is determined: a piece goods mean value (SKij, j=l...m), a smallest piece goods value (SK2j, j=l...m), a largest piece goods value (SK3j, j=l...m), a piece goods measurement duration value (SK4j, j=l...m), an article measurement time deviation value (SK5j, j=l...m), an article measurement deviation value (SK6j, j=l...m), an article curve integral value (SK7j, j=l...m), an article - Expected value (SK8j, j=l...m).

7. The method according to any one of claims 1 to 6, characterized in that for a machining operation (Bj, j=l...m) its piece goods parameter (SKij, i=l...n, j=l... .m) in the seventh step (V7) of the process is automatically classified into the good part class (GCj, j=l...m), the tool ( 12j, , j=l...m) is not worn out; and for a processing operation (Bj, j=l...m) its piece goods parameter (SKij, i=l...n, j=l...m) in the seventh step (V7) of the process automatically not can be classified into the good part class (GCj, j=l...m), the tool (12j, , j=l...m) required in this machining process (Bj, j=l...m) is worn out is.

8. The method according to any one of claims 1 to 7, characterized in that another piece of goods (l ^f ), which in the seventh step (V7) of the method in all processing operations (Bj, j = l ... m) automatically in the good part class (GCj, j=l...m) is classified, is a good part (G); and that a further piece good (1'), which in the seventh step (V7) of the method in all processing operations (Bj, j=l...m) automatically does not belong to the good part class (GCj, j=l...m ) can be classified, is a bad part (S).

9. The method according to any one of claims 1 to 8, characterized in that in the fourth step (V4) of the method, the determination of the good part class (GCj, j = l ... m) by monitored machine learning with a decision tree for graphical representation of hierarchically sequential decisions; and that with the decision tree the first piece good (1) is classified as a good part (G) by the good part characteristic value (GKij, i=l...n, j=l...m).

10. Machine tool (10) for carrying out the method for machining a workpiece (0, 0') according to one of claims 1 to 9, characterized in that the machine tool (10) has a tool holder (11), several required tools (12j, j=l...m) and a tool slide (13), which tool holder (11) holds the required tools (12j, j=l...m) simultaneously in different machining positions (p=l...m). ; and that the required tools (12j, j=l...m) are used to machine the workpiece (0, 0') in a chronological sequence of several machining processes (Bj, j=1...m).

11. Machine tool (10) according to claim 10, characterized in that the machine tool (10) has a piezoelectric force transducer (14), which piezoelectric force transducer (14) measures the tool force (F) in each machining process (Bj, j=l ...m) as a time sequence of measurement signals (Sj, j=l...m) in the form of electrical polarization charges; and that the piezoelectric force transducer (14) is a multi-component is a force transducer that measures a main cutting force component (Fx), a feed force component (Fy) and a passive force component (Fz) of the tool force (F).

12. Machine tool (10) according to claim 10, characterized in that the machine tool (10) has a piezoelectric force transducer (14), which piezoelectric force transducer (14) measures the tool force (F) in each machining process (Bj, j=l ...m) as a time sequence of measurement signals (Sj, j=l...m) in the form of electrical polarization charges; and that the piezoelectric force transducer (14) is a one-component force transducer which measures a main cutting force component (Fx) of the tool force (F).

13. Evaluation unit (30) for carrying out the method for machining a workpiece (0, 0 ') according to any one of claims 1 to 9, characterized in that the

Evaluation unit (30) has at least one computer (301); that the computer (301) has at least one data processor and at least one data memory; that at least one computer program (C) is stored in the data memory and can be loaded into the data processor; that the computer program (C) loaded into the data processor causes the computer (301) to carry out the third step (V3) of the method, the fourth step V4) of the method, the sixth step (V6) of the method and the seventh step (V7) of the to carry out the procedure.

14. Evaluation unit (30) according to claim 13, characterized in that the evaluation unit (30) has at least one converter unit (302); that a force transducer (14) measures the tool force (F) in each machining process (Bj, j=l...m) as a time sequence of measurement signals (Sj, j=l...m); that the force transducer (14) transmits the time sequence of measurement signals (Sj, j=l...m) to the converter unit (302); and that the converter unit (302) in the first step (VI) of the method and in the fifth step (V5) of the method for each processing operation (Bj, j=l...m) the temporal sequence of measurement signals (Sj, j =l...m) in the form of electrical polarization charges automatically into electrical DC voltages and automatically digitized into a chronological sequence of measured values (Xj, j=l...m).