EP4289555A1 - Double or single side machining machine and method for operating a double or single side machining machine - Google Patents

Abstract

The invention relates to a double- or single-sided processing machine with a preferably annular first working disk and a preferably annular counter-bearing element, wherein the first working disk and the counter-bearing element can be driven to rotate relative to one another by means of a rotary drive, and a preferably annular one between the first working disk and the counter-bearing element Working gap is formed for double-sided or single-sided processing of flat workpieces, preferably wafers, wherein the double- or single-sided processing machine comprises a plurality of sensors which, during operation of the double- or single-sided processing machine, provide measurement data on machine and/or processing parameters of the double - or single-side processing machine, a control device being provided which receives the measurement data recorded by the sensors during operation of the double- or single-side processing machine, the control device comprising an artificial neural network which is designed to generate a state vector from the measurement data of the double or single side processing machine and compare it with at least one target state vector. The invention also relates to a method for operating a double- or single-side processing machine.

Description

The invention relates to a double- or single-sided processing machine with a preferably annular first working disk and a preferably annular counter-bearing element, wherein the first working disk and the counter-bearing element can be driven to rotate relative to one another by means of a rotary drive, and a preferably annular one between the first working disk and the counter-bearing element Working gap is formed for double-sided or single-sided processing of flat workpieces, preferably wafers, wherein the double- or single-sided processing machine comprises a plurality of sensors which, during operation of the double- or single-sided processing machine, provide measurement data on machine and/or processing parameters of the double - or single-side processing machine. The invention also relates to a method for operating such a double- or single-side processing machine.

For example, in double-side polishing machines, flat workpieces, such as wafers, are polished between preferably ring-shaped work disks. A preferably annular working gap is arranged between the working disks, in which the flat workpieces, for example wafers, are held during processing. For this purpose, so-called rotor disks are usually arranged in the working gap with recesses in which the workpieces are mounted in a floating manner. The working discs are driven to rotate relative to one another for processing by means of a rotary drive and the rotor discs also rotate in the working gap via usually an external toothing of the rotor discs, which engages in a corresponding toothing of pin rings. As a result, the workpieces are conveyed through the working gap along cycloid paths during machining. During double-side polishing, a polishing agent, a so-called slurry, is added to the working gap, which ensures the abrasive processing. The In double-sided polishing machines, work disks also regularly have polishing cloths, so-called polishing pads, on the surfaces delimiting the working gap.

The aim of machining is to have the finished workpieces as plane-parallel as possible. The working gap geometry is crucial for this. Out of DE 10 2006 037 490 B4 is a double-side processing machine known with means for generating a global deformation of one of the working disks. In particular, the upper working disk can be deformed between a globally concave and a globally convex shape. With such a global deformation in the radial direction, the concave or convex shape of the working disk only arises over the entire diameter of the working disk. The annular surface of the preferably annular working disk delimiting the working gap remains flat in itself, but opposite ring sections of the annular surface are deformed relative to one another, so that overall a concave or convex shape results.

Out of DE 10 2016 102 223 A1 A double-side processing machine is also known with means for generating a local deformation of one of the working disks, in particular between a locally convex and a locally concave shape. With such a local deformation, the convex or concave shape results in the radial direction between the inner and outer edges of the, for example, annular working disk. In contrast to a global deformation, with a local deformation the ring sections themselves are concave or convex deformed.

The two aforementioned configurations can be combined in a double-side processing machine. In this way, a wide variety of working gap geometries can be created. For example, if the polishing cloths are partially worn out or if the temperatures of the components defining the working gap change, it can always be as plane-parallel as possible Processing of the workpieces or a setting of the working gap that is preferred for the workpiece quality, be it parallel or not, can be ensured.

The geometry of the working gap has a decisive influence on the shape and flatness of the machined workpiece. In addition to the geometry of the working gap, the machining result is influenced by a variety of other machine and/or machining parameters, for example the temperature of various components of the machine, the thickness and possible wear of a work surface, such as a polishing cloth, the speed of the work disk rotated relative to one another and/or the counter-bearing element and rotor disks rotatably mounted in the working gap, or for example a load between the first working disk and the counter-bearing element.

It is known to monitor such machine and/or processing parameters during operation of the double- or single-side processing machine using sensors. It is also known, as further processing parameters, to detect the shape and thickness of the flat workpieces, for example wafers, processed in the working gap using appropriate sensors. From the large number of these machine and processing parameters, a suitable parameter window must be found for operating the double or single-side processing machine, initially as part of setting up the double or single-side processing machine before starting to process the flat workpieces. The double or single-sided processing machine must also be adjusted to the boundary conditions prevailing at the respective location, for example the type of work surface, such as polishing cloths, if necessary a polishing agent and other parameter specifications of an operator. In the subsequent production operation of the double or single-sided processing machine, the process must be monitored using the sensors. Deviations from the specified target values, for example The GBIR or SFQR value of processed wafers can be detected early and, if necessary, corrective intervention can be made in the processing process.

Not least due to a large number of different machining processes, the measurement results from the sensors must be interpreted by expert personnel in order to reach the right conclusions for adapting production operations. Such expert personnel are not available at the site of every double or single-sided processing machine. This can lead to adverse effects on the production process. In addition, production operations are often only adjusted with a considerable time delay after any harmful parameter deviations occur. One reason for this is that, due to the large number of machine and processing parameters that influence the production process, it is difficult, even for expert personnel, to recognize a relevant deviation in the measured parameters at an early stage. This often only happens after the finished, machined workpieces have been measured. If an undesirable deviation is then discovered in the production process, a significant amount of rejects will arise in the meantime.

Based on the explained prior art, the invention is based on the object of providing a double- or single-side processing machine and a method for operating a double- or single-side processing machine with which the production process of the double- or single-side processing machine can be carried out more quickly and reliably Minimization of rejects can be monitored.

The invention solves the problem through the independent claims 1 and 9. Advantageous refinements can be found in the dependent claims, the description and the figures.

For a double- or single-side processing machine of the type mentioned, the invention solves the problem in that a control device is provided which receives the measurement data recorded by the sensors during operation of the double- or single-side processing machine, the control device being an artificial neural network includes, which is designed to create a state vector of the double- or single-side processing machine from the measurement data and to compare this with at least one target state vector.

For a method of the type mentioned at the outset, the invention solves the problem in that the artificial neural network is trained by entering a large number of target state vectors leading to an acceptable machining result of flat workpieces.

The double or single-side processing machine according to the invention can in particular be a double or single-side polishing machine. The double or single side processing machine can also be a double or single side lapping machine or double or single side grinding machine. The double or single-sided processing machine has a preferably annular first working disk and a preferably annular counter-bearing element. In the case of a single-sided processing machine, the counter-bearing element can be designed, for example, as a simple weight or pressure cylinder. The counter-bearing element can preferably be a preferably annular second working disk. The first working disk and the counter-bearing element can be driven to rotate relative to one another and a preferably annular working gap for processing flat workpieces, such as wafers, is formed between the first working disk and the counter-bearing element. In particular if it is a double or single-sided polishing machine, at least the first working disk, preferably also the counter-bearing element or the second working disk, can have a polishing coating (polishing pad) on its surface(s) delimiting the working gap. exhibit. During processing, a process medium, for example a polishing agent, in particular a polishing liquid (slurry), can also be added to the working gap in a manner known per se. The work disks can also be provided with temperature control channels through which a temperature control fluid, for example cooling water, is passed during operation to control the temperature of the work disk(s).

The double or single-sided processing machine is used in particular for plane-parallel processing of flat workpieces. For machining, the workpieces can be held floating in recesses of rotor disks arranged in the working gap in a manner known per se. The first working disk and the counter-bearing element are driven to rotate relative to one another during operation, for example via a corresponding drive shaft and at least one drive motor. It is possible that only one of the first working disk and the counter bearing element is driven in rotation. However, both the first working disk and the counter-bearing element can also be driven in rotation, usually in opposite directions. For example, in a double-side processing machine, the rotor disks can also be moved in a rotating manner through the working gap in the course of the relative rotation between the first working disk and the counter bearing element using suitable kinematics, so that workpieces arranged in the recesses of the rotor disks describe cycloid paths in the working gap. For example, the rotor disks can have a toothing on their outer edge, which engages with an associated toothing of pin rings. Such machines form what is known as planetary kinematics.

The first working disk and/or the counter bearing element can each be held by a carrier disk. Like the first working disk and the counter-bearing element, the carrier disks can also be annular or at least have annular carrier sections.

In a manner known per se, according to the invention, sensors, in particular suitable measuring devices, record measurement data relating to machine and/or processing parameters of the double- or single-side processing machine during operation of the double- or single-side processing machine. This can in particular be the machine and/or processing parameters mentioned at the beginning. The sensors record the measurement data in particular at certain intervals or continuously. The measurement data characterize the operating and machine parameters of the double or single-sided processing machine and thus the production process. The measurement data recorded by the sensors are also fed, in particular, at certain intervals or continuously to a control device of the double- or single-side processing machine according to the invention. The machine and/or processing parameters can be recorded in real time. This also applies to the transfer of the measurement data to the control device and the processing of the measurement data explained below. The recorded measurement data can also be stored in a data memory and passed on from there to the control device, for example in real time or delayed.

To process the measurement data, the control device according to the invention comprises an artificial neural network, which creates a state vector of the double- or single-side processing machine from the received measurement data. The state vector is composed of the current measurement data from the sensors or is formed from this current measurement data. The state vector characterizes the double- or single-sided processing machine, and in particular the current production process. The artificial neural network compares this state vector with at least one, preferably a plurality, in particular a group, of target state vectors. The target state vectors are given to the artificial neural network according to the method according to the invention as part of training as state vectors for an acceptable processing result when processing Workpieces in the double or single side processing machine. The target state vectors can be defined with different objectives, for example aimed at certain quality parameters (e.g. GBIR and/or SFQR) and/or production throughput or other parameters. By comparing the state vector created from the current measurement data with the target state vectors available to the artificial neural network, it can determine whether the current state vector matches one of the target state vectors trained as acceptable or not. If it is determined that the recorded state vector does not sufficiently match any of the acceptable target state vectors, countermeasures can be taken, for example, in the event of a relevant deviation from the acceptable target state vectors. For example, the production process can be intervened by adjusting machine and/or processing parameters. Of course, defined tolerances can be specified for the comparison, within which a detected small deviation from the target state vectors is classified as acceptable. By adjusting machine and/or processing parameters based on the comparison, the production process can be influenced in such a way that the currently created state vector (again) sufficiently matches at least one target state vector.

Unlike an operator, an artificial neural network can very quickly create a state vector from a large number of measurement data and thus machine and/or processing parameters and can also compare this very quickly with at least one, preferably a large number of target state vectors. This means that an impermissible deviation of the production process from an acceptable process can be detected quickly and reliably, especially if there are no sufficiently trained or experienced personnel available at the production site of the double- or single-side processing machine. The invention takes advantage of the fact that with an optimal production process the measurable Machine and/or processing parameters have a fixed relationship to one another. An artificial neural network that is trained with the machine and/or machining parameters of an optimal process can therefore quickly and reliably detect deviations of the current process from the optimal process. The artificial neural network forms an anomaly detector that identifies an unacceptable deviation (anomaly) in the production process. Process optimization is therefore possible even when starting the production process after an initial setup process much faster and with a significantly smaller number of test production processes. In the best case, only a single test production attempt is required, which does not require any external downstream measurement of the machined workpieces. Monitoring the production process, even during the start of production, is easier and faster and minimizes waste. In particular, the production of workpieces in the double- or single-side processing machine that do not lie within the desired tolerances can be reduced or, at best, completely avoided.

According to one embodiment, the control device can be designed to issue a warning message if the created state vector deviates from the at least one target state vector. The warning message can be issued to an operator of the double- or single-side processing machine, for example via a user interface of the double- or single-side processing machine. In the simplest case, the operator receives a warning message that machine and/or processing parameters deviate impermissibly from values acceptable for an optimal production process. On this basis, the operator can intervene manually in the process, in particular specifically adjusting machine and/or processing parameters, so that the state vector formed from the current measurement data (again) corresponds to at least one target state vector.

In a further variant, the warning message can already include an adjustment suggestion for adjusting certain machine and/or processing parameters. This adjustment suggestion can be issued by the control device on the basis of an adjustment rule stored in the control device. Such an adjustment rule can have been created in advance by an operator for the double- or single-side processing machine. The operator can then evaluate the adjustment suggestion and, if necessary, carry it out. By combining determined deviation values between the state vector and the at least one target state vector with formalized causalities of the machine and/or processing parameters of the double- or single-side processing machine, the control device can automatically create suggestions for changing machine and/or processing parameters.

According to a further embodiment, the control device can further comprise a control device which is designed to control the double- or single-side processing machine, in particular machine and/or operating parameters, in the event of a deviation of the created state vector from the at least one target state vector as determined by the comparison Double or single-sided processing machine to be controlled so that the created state vector matches at least one target state vector. For this purpose, the control device can in particular control actuators for influencing the machine and/or processing parameters. The control device achieves further automation by independently controlling the double- or single-side processing machine based on the comparison carried out, so that the state vector created from the current measurement data (again) corresponds to at least one of the target state vectors. The control device can be integrated into the control device.

The control device can be designed to control the machine and/or operating parameters of the double- or single-side processing machine based on an adaptation rule stored in the control device. The adaptation rule can in particular specify certain control regulations for certain determined deviations of the state vector to the control device. Again, the adaptation rule can, for example, have been created in advance by an operator. On this basis, automated control based on control specifications stored in advance in the form of the adjustment rule is possible, in particular without intervention by an operator.

According to a further embodiment, a further artificial neural network can be provided, which is designed to use machine learning to evaluate the measurement data for the machine and / or processing parameters and, based on the evaluation, the double- or single-sided processing machine, in particular machine and /or operating parameters of the double- or single-side processing machine, and/or to create and/or change an adaptation rule stored in a control device. This further artificial neural network can be a further artificial neural network, in addition to the aforementioned first artificial neural network forming the anomaly detector. However, it would also be conceivable for the further artificial neural network to be designed to be integrated with the aforementioned artificial neural network forming the anomaly detector. The control device can be integrated into the further artificial neural network.

The additional artificial neural network provided can in particular comprise a so-called learning classifier system (LCS/learning classifier system), i.e. an artificial intelligence system. Such systems are based on defined if-then relationships and can include machine and/or processing parameters of the double- or single-side processing machine Dependency on anomaly values, i.e. deviations between the current state vector and the at least one target state vector detected by the (first) artificial neural network. The LCS creates output data from input data and rules. The control device can also include a memory in which machine and/or machining parameters obtained in the past, including data on workpieces to be machined, are stored. The stored data can be made available to the artificial neural network, in particular the LCS, which takes this data into account when evaluating the measurement data and the resulting control data for the double- or single-side processing machine. The artificial neural network, which is preferably designed as an LCS, can then recognize during a production process the probability with which the machining result of the workpieces, for example characteristic values such as GBIR, SFQR, deviate from predetermined target values. On this basis, this artificial neural network can intervene during the production process or at the latest in a subsequent production process by controlling, for example, actuators for certain machine and/or processing parameters in order to avoid any waste. The artificial neural network using machine learning can also improve an adaptation rule initially created by an operator, for example, based on further experience from production processes. For this purpose, the artificial neural network can change an adaptation rule stored in a control device. It would also be conceivable for this adaptation rule to be created by the artificial neural network and then, if necessary, optimized based on further process data. The aforementioned design makes it possible to automate the production process to the greatest possible extent without the intervention of operators.

According to a further embodiment, it can be provided that the sensors include measuring devices for measuring the working gap, in particular the shape and/or width of the working gap, further in particular a distance between the first working disk and the counter-bearing element, and/or for measuring a temperature of the first working disk and/or the counter-bearing element and/or of further machine components of the double- or single-sided processing machine and/or for measuring a temperature and/or a flow rate of a processing means fed into the working gap for processing the workpieces, and/or for measuring a speed of the first working disk and/or the counter-bearing element and/or of rotor disks rotatably mounted in the working gap and/or for measuring a load between the first working disk and the counter bearing element and/or for measuring a speed and/or a torque and/or a temperature of the rotary drive and/or for measuring a pressure and/or a force of means for generating a deformation of the first working disk and/or of the counter-bearing element and/or for measuring the thickness of a working surface of the first work disk and/or of the counter-bearing element and/or for measuring the thickness and/or shape of workpieces processed in the double- or single-side processing machine. The measuring devices mentioned can be present together or in any combination with one another. Processing agents can be, for example, polishing agents, in particular polishing liquids such as slurry. The measuring devices mentioned record machine and processing parameters relevant to the production process, including, for example, environmental data from the double-sided or single-sided processing machine.

If the double-sided or single-sided processing machine, in particular machine and/or operating parameters of the double-sided or single-sided processing machine, are controlled based on a deviation of the created state vector from the at least one target state vector determined by the comparison, this can in particular be controlled of actuators to influence the working gap, in particular the shape and/or width of the Working gap, further in particular a distance between the first working disk and the counter-bearing element, and/or for influencing a temperature of the first working disk and/or the counter-bearing element and/or of further machine components of the double- or single-side processing machine and/or for influencing a temperature and /or a flow rate of a processing agent supplied into the working gap for processing the workpieces, and/or for influencing a speed of the first working disk and/or the counter-bearing element and/or of rotor disks rotatably mounted in the working gap and/or for influencing a load between the first working disk and counter bearing element and/or for influencing a speed and/or a torque and/or a temperature of the rotary drive and/or for influencing a pressure and/or a force of means for generating a deformation of the first working disk and/or the counter bearing element and/or for influencing the thickness of a working surface of the first working disk and/or the counter-bearing element and/or for influencing the thickness and/or shape of workpieces processed in the double- or single-sided processing machine. The mentioned influences or controls of the actuators can take place together or in any combination with one another. Processing agents can be, for example, polishing agents, in particular polishing liquids such as slurry. The actuators to be controlled therefore influence machine and processing parameters relevant to the production process, including, for example, environmental data of the double- or single-sided processing machine.

According to a further embodiment, it can be provided that the counter-bearing element is formed by a preferably annular second working disk, the first and second working disks being arranged coaxially to one another and being able to be driven in rotation relative to one another by the rotary drive, with the working gap between the working disks for double-sided or one-sided processing flat workpieces is formed.

The invention also relates to a system comprising at least two double-sided or single-sided processing machines according to the invention, wherein a higher-level artificial neural network is also provided, which is connected to the artificial neural networks of the at least two double-sided or single-sided processing machines, the higher-level artificial neural network Network is designed to, based on data obtained from the artificial neural networks of the at least two double- or single-sided processing machines, at least one artificial neural network of the at least two double- or single-sided processing machines by inputting state vectors leading to an acceptable processing result of flat workpieces to train.

In this embodiment, a system of at least two, in particular more than two, double- or single-side processing machines according to the invention is provided. Furthermore, a higher-level artificial neural network is provided, which is connected to the artificial neural networks of the at least two double- or single-sided processing machines. The higher-level artificial neural network is designed to train at least one artificial neural network of the at least two double-sided or single-sided processing machines based on the data received from the artificial neural networks of the at least two double-sided or single-sided processing machines. The higher-level artificial neural network therefore forms a higher-level structure into which similar double- or single-side processing machines can be integrated. A system-wide memory can then also be provided, which receives data from all double- or single-side processing machines in the system and also passes this data on to the higher-level artificial neural network. In this way, the individual double- or single-side processing machines can be optimized, if necessary, taking into account the data stored in the memory of the system with mutual use of individual data from the double or single-sided processing machines of the system. The aforementioned design allows advantageous effects to be achieved, for example with regard to production planning, fleet management or maintenance prediction (predictive maintenance).

The double- or single-side processing machine according to the invention can be designed to carry out the method according to the invention. Accordingly, the method according to the invention can be carried out with the double- or single-side processing machine according to the invention.

As already explained, in the method according to the invention, the artificial neural network is trained by inputting a large number of state vectors leading to an acceptable machining result of flat workpieces. The training can be carried out by an operator carrying out production processes with the double- or single-side processing machine with different machine and/or processing parameters and, depending on the processing result, specifying the respective machine and/or processing parameters to the artificial neural network the production process led to an acceptable processing result. In this case, the associated machine and/or processing parameters are stored as a target state vector in the artificial neural network. This initial training usually takes place before regular processing of flat workpieces begins with the double- or single-side processing machine.

It is also possible for the artificial neural network trained in this way to be further trained during operation of the double- or single-side processing machine by entering further target state vectors leading to an acceptable processing result of flat workpieces. Through this further training in the course Production processes with double or single-sided processing machines result in further optimization of the machine and/or processing parameters.

According to a further embodiment, during operation of the double- or single-sided processing machine with the trained artificial neural network, a further artificial neural network can be trained by entering a large number of target state vectors leading to an acceptable processing result of flat workpieces. This additional artificial neural network can be untrained or already (pre-)trained. For example, the further artificial neural network can be a copy of the trained artificial neural network and can be further trained on this basis. This can be useful, for example, if the trained artificial neural network is a generic neural network that is trained for a specific type of double or single-sided processing machine, but has not yet been specialized for a specific double or single-sided processing machine, in particular with regard to the individual processing parameters on site. This allows a specialized version of the trained artificial neural network to be generated, which can ultimately replace the trained artificial neural network. A possible application is that double-sided or single-sided processing machines are delivered with a trained artificial neural network, with the training taking place on the basis of experiments or laboratory data from a manufacturer of the double-sided or single-sided processing machine and then with the further recent neural network Further specialization takes place on the customer's individual manufacturing process. This requires less understanding of the production process at the installation site of the double or single side processing machine.

Figur 1

einen Teil einer erfindungsgemäßen Doppel- oder Einseiten-Bearbeitungsmaschine in einer Schnittansicht in einem ersten Betriebszustand,

Figur 2

die Ansicht aus Figur 1 in einem zweiten Betriebszustand,

Figur 3

die Ansicht aus Figur 1 in einem dritten Betriebszustand,

Figur 4

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem ersten Ausführungsbeispiel,

Figur 5

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem weiteren Ausführungsbeispiel,

Figur 6

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem weiteren Ausführungsbeispiel,

Figur 7

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem weiteren Ausführungsbeispiel,

Figur 8

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem weiteren Ausführungsbeispiel,

Figur 9

eine schematische Darstellung der Funktion der erfindungsgemäßen Doppelseiten-Bearbeitungsmaschine nach einem weiteren Ausführungsbeispiel, und

Figur 10

ein erfindungsgemäßes System in einer schematischen Darstellung.

Exemplary embodiments of the invention are explained in more detail below with reference to figures. It shows schematically:

Figure 1

a part of a double- or single-side processing machine according to the invention in a sectional view in a first operating state,

Figure 2

the view Figure 1 in a second operating state,

Figure 3

the view Figure 1 in a third operating state,

Figure 4

a schematic representation of the function of the double-side processing machine according to the invention according to a first exemplary embodiment,

Figure 5

a schematic representation of the function of the double-side processing machine according to the invention according to a further exemplary embodiment,

Figure 6

a schematic representation of the function of the double-side processing machine according to the invention according to a further exemplary embodiment,

Figure 7

a schematic representation of the function of the double-side processing machine according to the invention according to a further exemplary embodiment,

Figure 8

a schematic representation of the function of the double-side processing machine according to the invention according to a further exemplary embodiment,

Figure 9

a schematic representation of the function of the double-side processing machine according to the invention according to a further exemplary embodiment, and

Figure 10

a system according to the invention in a schematic representation.

Unless otherwise stated, the same reference numbers designate the same objects in the figures.

The ones in the Figures 1 to 3 The double-side processing machine, shown only as an example, has an annular upper carrier disk 10 and a likewise annular lower carrier disk 12. A first, annular upper working disk 14 is attached to the upper carrier disk 10 and a second, also annular working disk 16 is attached to the lower carrier disk 12. A likewise annular working gap 18 is formed between the annular working disks 14, 16, in which flat workpieces, for example wafers, are processed on both sides during operation. The double-side processing machine can be, for example, a polishing machine, a lapping machine or a grinding machine.

The upper carrier disk 10 and with it the upper working disk 14 and / or the lower carrier disk 12 and with it the lower working disk 16 can be driven in rotation relative to one another by a suitable drive device, comprising, for example, an upper drive shaft and / or a lower drive shaft and at least one drive motor become. The drive device is known per se and is not shown in more detail for reasons of clarity. In a manner also known per se, the workpieces to be machined can be held floating in rotor disks in the working gap 18. Suitable kinematics, for example planetary kinematics, can ensure that The rotor disks also rotate through the working gap 18 in the course of the relative rotation of the carrier disks 10, 12 or working disks 14, 16. In the upper working disk 14 or the upper carrier disk 10 and possibly also the lower working disk 16 or the lower carrier disk 12, temperature control channels can be formed, through which a temperature control fluid, for example a temperature control liquid such as water, can be passed during operation. This is also known per se and is not shown in more detail.

The ones in the Figures 1 to 3 The double-side processing machine shown also includes known distance measuring devices as sensors. The sensors can, for example, work optically or electromagnetically (e.g. eddy current sensors). In the example shown, for example, three distance measuring devices 20, 22, 24 are provided, which measure the distance between the upper working disk 14 and the lower working disk 16 at three radially spaced positions of the working gap 18, as in Figure 1 illustrated by arrows. As can be seen, the distance measuring device 20 measures the distance between the upper working disk 14 and the lower working disk 16 in the area of the radially outer edge of the working gap 18. The distance measuring device 24 measures the distance between the upper working disk 14 and the lower working disk 16 in the area of the radial inner edge of the working gap 18. The distance measuring device 22 measures the distance between the upper working disk 14 and the lower working disk 16 in the middle of the working gap 18.

In the Figures 2 and 3 the distance measuring devices 20, 22, 24 are not shown for reasons of clarity. The measurement data from the distance measuring devices 20, 22, 24 are available at a control device 34.

In the present case, the lower working disk 16 is only attached to the lower carrier disk 12 in the area of its outer edge and in the area of its inner edge, For example, each screwed along a partial circle, as in Figure 1 illustrated at reference numbers 26 and 28. However, between these fastening locations 26 and 28, the lower working disk 16 is not attached to the lower carrier disk 12. Rather, there is an annular pressure volume 30 between these fastening locations 26, 28 between the lower carrier disk 12 and the lower working disk 16. The pressure volume 30 is connected via a dynamic pressure line 32 to a pressure fluid reservoir (not shown in detail in the figures), for example a liquid reservoir, in particular a water reservoir. tied together. A pump and a control valve can be arranged in the dynamic pressure line 32, which can be controlled by the control device 34. In this way, a desired pressure can be built up in the pressure volume 30 by fluid introduced into the pressure volume 30, which then acts on the lower working disk 16. The pressure prevailing in the pressure volume 30 can be measured via a pressure measuring device (not shown). The measurement data from the pressure measuring device as a further sensor can also be applied to the control device 34, so that the control device 34 can set a predetermined pressure in the pressure volume 30.

Due to its freedom of movement between the fastening locations 26, 28, the lower working disk 16 can be locally brought into a convex shape by setting a sufficiently high pressure in the pressure volume 30, as in Figure 2 indicated by dashed lines at reference number 36. If you go in the operating state of the Figure 1 , in which the lower working disk 16 has a flat shape, from a pressure po in the pressure volume 30, the in Figure 2 Convex deformation of the lower working disk 16 shown at 36 can be achieved by setting a pressure p ₁ >p ₀ . On the other hand, by setting a pressure p ₂ <p ₀ in the printing volume 30, a local concave deformation of the lower working disk 16 can be achieved, as in Figure 3 illustrated in dashed lines at reference number 38.

It can be seen that the lower working disk 16, viewed in the radial direction, has a locally convex shape between its inner edge, in the area of the attachment location 26, and its outer edge, in the area of the attachment location 28 ( Figure 2 ) or a locally concave shape ( Figure 3 ) can accept.

In addition to this local radial deformation of the lower working disk 16, means for global deformation of the upper working disk 14 can be provided. These means can be designed as explained above or in the DE 10 2006 037 490 B4 described. In this case, the upper carrier disk 10 and with it the upper working disk 14 attached to it are globally deformed, so that a globally concave or globally convex shape of the working surface of the upper working disk 14 results over the entire cross section of the upper working disk 14. However, between its radially inner edge and its radially outer edge, the upper working disk 14 can remain flat or can be locally deformed by the pressure volume 30 in the manner explained above. The means for adjusting the shape of the upper working disk 14 can also be controlled by the control device 34.

The distance measuring devices 20, 22, 24 form sensors which, during operation of the double-side processing machine, record measurement data on machine and/or processing parameters of the double-side processing machine, in the present case in particular the thickness and geometry of the working gap 18. The double-side processing machine preferably comprises a plurality additional sensors with corresponding additional measuring devices. These can in particular be measuring devices of the type explained above. These measuring devices record further machine and/or processing parameters during operation of the double-side processing machine.

The measurement data recorded by the sensors are sent to the control device 34. From these measurement data, the control device 34 creates a state vector of the double-side processing machine by means of an artificial neural network 34 integrated therein and compares this with at least one target state vector, preferably a group of target state vectors, which were assigned to an acceptable production process as part of training .

Based on Figure 4 The training of the artificial neural network 34 will be explained in more detail. In Figure 4 the double-side processing machine according to the invention is shown at reference number 40. Unprocessed workpieces 42, for example unprocessed wafers, are fed to it for processing and finished workpieces 44, in particular processed wafers 44, are output by the double-side processing machine 40. A data memory 46 is provided, to which, for example, measurement data on machine and processing parameters recorded by the sensors are supplied, with this data being made available to an operator 48, as in Figure 4 illustrated at 50. Furthermore, measurement data relating to, for example, the geometry of the machined workpieces are supplied to the data memory 46 as further machine and/or processing parameters, with these data also being supplied to the operator 48, as in Figure 4 illustrated at 52. Finally, external environmental data is also available to the data storage, as illustrated at 54. This external environmental data can also be sent to the operator 48. On this basis, the operator 48 carries out an assessment of the production process underlying the respective data to determine whether the processing result is acceptable. The operator 48 provides this evaluation to the artificial neural network 34 of the control device 34, as in Figure 4 shown at 56. The corresponding state vectors are stored as target state vectors by the artificial neural network 34.

In Figure 5 shows how the double-side processing machine can be operated on this basis. In this case, the process data on the machine and/or processing parameters are fed directly to the artificial neural network 34 of the control device 34, as in Figure 5 illustrated at 58. The artificial neural network 34 creates a state vector from the measurement data obtained for the machine and/or processing parameters and compares it with the stored target state vectors. If an impermissible deviation or non-conformity is detected, the control device 34 issues a corresponding warning message to the operator 48, as in Figure 5 shown at 60. On this basis and, if necessary, taking into account the measurement data of the processed workpieces 44 provided via 52, the operator 48 can control the double-side processing machine 40, in particular actuators for influencing machine and / or operating parameters, as in Figure 5 shown at 62 to bring the continuously monitored and created state vector into agreement with at least one target state vector. In this case, the operator 48 decides on the consequences of evaluating the received data. The operator 48 is supported by the control device 34 as an anomaly detector.

In Figure 6 is a further automated variant of the in Figure 5 procedure shown. In this embodiment, the control device 34, in particular its artificial neural network 34, further comprises a control device 64 connected thereto, as in Figure 6 shown at 66. If the comparison determines a deviation of the created state vector from the at least one target state vector, a control intervention is carried out by the control device 64 on actuators of the double-side processing machine 40, in particular without intervention by the operator 48. This results in an adaptation of the recorded machine and/or processing parameters the double-side processing machine 40 causes as in Figure 6 illustrated at 68. All the associated data can be stored in the data memory 46. The control and regulation device 34, 64, which is designed to be integrated, for example, can control the machine and/or operating parameters of the double-side processing machine 40 on the basis of an adaptation rule stored, for example, in the control device 64. This can, for example, contain specific control instructions created by an operator 48 for certain identified deviations in the machine and/or processing parameters, according to which the open-loop and closed-loop control device 34, 64 controls actuators.

In Figure 7 is another embodiment of the too Figure 6 the procedure explained. In this embodiment, the operator 48 is still involved. This also receives the process data on the recorded machine and processing parameters, as in Figure 7 shown at 70 and also the process data for the machined workpieces, as shown at 52. Finally, the operator 48 also receives the control commands issued by the control device 34, as in Figure 7 shown at 72. On this basis, the operator 48 can monitor the control carried out in each case and, if necessary, adjust the control of the control device 64 in a suitable manner, as in Figure 7 shown at 74.

Figure 8 represents a further embodiment of a possible training of artificial neural networks as anomaly detectors. This is based on the control device 34 with an already pre-trained artificial neural network 34, trained for example as above Figure 4 explained. This control device 34 sends any deviations or anomaly data to the operator 48, as in Figure 8 shown at 60 and closed at the top Figure 5 explained. On this basis, the operator 48 can train a further artificial neural network 76 by giving this further artificial neural network 76 (further) target state vectors acceptable during operation of the double-side processing machine 40 Machine and/or processing parameters of the double-side processing machine 40 are supplied, as in Figure 8 shown at 78. The further artificial neural network 76 can be an untrained artificial neural network 76. However, it can also be an already (pre-) trained artificial neural network 76, for example a duplicate of the neural network 34 of the control device 34. On this basis, for example for the generic type of the double-side processing machine 40 (before -) trained artificial neural network 34 of the control device 34 in the beginning of production operation, a specialized training of the further artificial neural network 76 can be trained for the respective individual process parameters of the application of the double-side processing machine 40. It is possible that after this training has been completed, the further artificial neural network 76 replaces the previously trained artificial neural network 34 of the control device 34.

Based on Figures 9 and 10 A further embodiment of the invention will be explained, comprising in particular a further artificial neural network 86, designed for machine learning. It can be a Learning Classifier System (LCS), i.e. an artificial intelligence system. At the in Figure 9 In the embodiment shown, measurement data from the sensors for machine and/or processing parameters are fed to the data memory 46 on the one hand and to the control device 34 on the other hand, as in Figure 9 shown at 80. Workpiece data, in particular measurement data on the geometry of the machined workpieces, are also supplied to both the data memory 46 and the control device 34, as in Figure 9 shown at 82. The control device 34 is also in communication with the data memory 46, as in Figure 9 shown at 84. In Figure 9 Another artificial neural network 86 is shown, which is also associated with the control device 34. This further artificial neural network 86 can also be combined with the artificial neural network 34 of the control device 34. This Another artificial neural network 86 is designed for machine learning and in particular forms a learning classifier system (LCS).

The measurement data on the geometry of the machined workpieces 44 are also sent to the LCS 86 via 82. If the control device 34, in particular its artificial neural network 34, detects an unacceptable deviation between the currently recorded state vector and the acceptable values of the machine and/or processing parameters stored as target state vectors during operation of the double-side processing machine 40, a corresponding anomaly signal is generated given to the LCS 86, as in Figure 9 shown at 88. The LCS 86 can also have measurement data from the past from the data memory 46 available. On this basis, the LCS 86 can independently make decisions about changing certain process parameters, in particular the control of actuators to influence the recorded machine and/or processing parameters, and control the actuators accordingly, as in Figure 9 shown at 90. In this way, the greatest possible automation and independence can be achieved.

In Figure 10 is a further embodiment of the in Figure 9 shown variant shown. In particular shows Figure 10 a system according to the invention with at least two double-side processing machines 40. Of course, the system could also include more than two double-side processing machines 40, which in Figure 10 is illustrated by three points. In Figure 10 For purposes of illustration, two systems 92i are shown in dashed blocks, each of which corresponds to the design in terms of its design and function Figure 9 can correspond. The systems 92i can also be designed differently, for example designed to achieve different goals, for example optimal wafer quality, maximum output, etc. They are connected to a common data memory 46 via 80, 84. In addition, the in figure 10 In the system shown, a higher-level artificial neural network 94 is provided, again comprising, for example, an LCS, which can also be connected to an operator 48. The higher-level artificial neural network 94 is also connected to the data memory 46, as shown at 96. In addition, the higher-level artificial neural network 96 receives the control commands executed by the LCS 86, as in Figure 10 shown at 98. On this basis, the higher-level LCS 94 can further optimize or specialize the LCS 86 of the systems 92i based on data from the systems 92i, for example by specifying collective or individual control regulations and / or target state vectors for the individual systems 92i.

Reference symbol list

10

upper carrier disk

12

lower carrier disk

14

upper working disc

16

lower work disk

16

Counter bearing element

18

working gap

20, 22, 24

Distance measuring device, sensors

26

Mounting location

28

Mounting location

30

Print volume

32

Back pressure line

34

Control device, artificial neural network

36, 38, 50

Arrow

52, 54, 56

Arrow

58,60, 62

Arrow

66, 68, 70

Arrow

72, 74, 78

Arrow

80, 82, 84

Arrow

88, 90, 96

Arrow

98

Arrow

40

Double side processing machine

42

unprocessed workpieces

44

processed workpieces

46

Data storage

48

Operator

64

Control device

76, 86, 94

artificial neural network

92i

Investments

Claims (14)

Double or single-sided processing machine with a preferably annular first working disk (14) and a preferably annular counter-bearing element (16), wherein the first working disk (14) and the counter-bearing element (16) can be driven in rotation relative to one another by means of a rotary drive, and between the The first working disk (14) and the counter bearing element (16) form a preferably annular working gap (18) for double-sided or single-sided processing of flat workpieces (42, 44), preferably wafers, the double- or single-sided processing machine having a plurality of sensors ( 20, 22, 24), which record measurement data on machine and/or processing parameters of the double- or single-side processing machine during operation of the double- or single-side processing machine, characterized in that a control device (34) is provided which is in Operation of the double or single side processing machine receives the measurement data recorded by the sensors (20, 22, 24), the control device (34) comprising an artificial neural network (34) which is designed to generate a state vector of the double from the measurement data - or single-side processing machine and compare it with at least one target state vector. Double or single-side processing machine according to claim 1, characterized in that the control device (34) is designed to issue a warning message if the created state vector deviates from the at least one target state vector. Double-sided or single-sided processing machine according to one of the preceding claims, characterized in that the control device (34) further comprises a control device (64) which is designed to operate at one through the Comparison determined deviation of the created state vector from the at least one target state vector to control the double or single-side processing machine, in particular machine and / or operating parameters of the double or single-side processing machine, so that the created state vector matches at least one target state vector . Double or single-side processing machine according to claim 3, characterized in that the control device (64) is integrated into the control device (34). Double- or single-side processing machine according to one of claims 3 or 4, characterized in that the control device (64) is designed to control the machine and/or operating parameters of the double- or single-side processing machine on the basis of a value in the control device (64). the stored adjustment rule. Double-sided or single-sided processing machine according to one of the preceding claims, characterized in that a further artificial neural network (86) is provided, which is designed to use machine learning to evaluate the measurement data on the machine and/or processing parameters and based on this the evaluation, to control the double or single-side processing machine, in particular machine and/or operating parameters of the double or single-side processing machine, and/or to create and/or change an adaptation rule stored in a control device (64). Double or single-side processing machine according to claim 6, characterized in that the control device (64) is integrated into the further artificial neural network (86). Double or single-sided processing machine according to one of the preceding claims, characterized in that the sensors (20, 22, 24) comprise measuring devices (20, 22, 24) for measuring the working gap (18), in particular a distance between the first working disk ( 14) and the counter-bearing element (16), and/or for measuring a temperature of the first working disk (14) and/or the counter-bearing element (16) and/or of further machine components of the double- or single-side processing machine and/or for measuring a temperature and/or a flow rate of a processing agent supplied into the working gap (18) for processing the workpieces (42, 44), and/or for measuring a speed of the first working disk (14) and/or of the counter bearing element (16) and/or of in rotor disks rotatably mounted in the working gap (18) and/or for measuring a load between the first working disk (14) and counter bearing element (16) and/or for measuring a speed and/or a torque and/or a temperature of the rotary drive and/or for measuring a pressure and/or a force of means for generating a deformation of the first working disk (14) and/or the counter-bearing element (16) and/or for measuring the thickness of a working surface of the first working disk (14) and/or the counter-bearing element (16) and/or for measuring the thickness and/or shape of workpieces (44) processed in the double- or single-side processing machine. Double or single-sided processing machine according to one of the preceding claims, characterized in that the counter-bearing element (16) is formed by a preferably annular second working disk (16), the first and second working disks (14, 16) being arranged coaxially to one another and through the Rotary drive can be driven to rotate relative to one another, the working gap (18) being formed between the working disks (14, 16) for double-sided or one-sided machining of flat workpieces (42, 44). Double or single-side processing machine according to one of the preceding claims, characterized in that it is designed to carry out the method according to one of claims 12 to 14. System comprising at least two double- or single-page processing machines according to one of the preceding claims, characterized in that a higher-level artificial neural network (94) is further provided, which is connected to the artificial neural networks (34, 86) of the at least two double- or single-pages -Processing machines is connected, the higher-level artificial neural network (94) being designed to generate at least one artificial neural network (34, 86) to train the at least two double- or single-sided processing machines by entering state vectors leading to an acceptable processing result of flat workpieces (42, 44). Method for operating a double- or single-sided processing machine according to one of the preceding claims, characterized in that the artificial neural network (34) is trained by inputting a plurality of target state vectors leading to an acceptable processing result of flat workpieces (42, 44). . Method according to claim 12, characterized in that the trained artificial neural network (34) is further trained during operation of the double- or single-sided processing machine by entering further target state vectors leading to an acceptable processing result of flat workpieces (42, 44). Method according to one of claims 12 or 13, characterized in that during operation of the double- or single-side processing machine with the trained artificial neural network (34), a further artificial neural network (76) is created by entering a plurality of flat Workpieces (42, 44) leading target state vectors is trained.

Images