Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
REDUCTION OF FRICTION WITHIN A MACHINE TOOL
Document Type and Number:
WIPO Patent Application WO/2021/197935
Kind Code:
A1
Abstract:
The invention relates to a computer-implemented method for reducing friction within a machine tool (MT), comprising the method steps: a) reading (S1) a plurality of surrogate models (SM) for approximating friction compensation within a given machine tool, wherein each surrogate model (SM1,..., SMm) is configured such that it assigns a friction compensation result value to a given friction compensation parameter set for reducing friction within a machine tool, and wherein a weighting factor (w1,...,wm) is assigned to each surrogate model (SM1,...,SMm), b) reading (S2) a friction compensation parameter set (CP), c) determining (S3) a friction compensation result value (CPR1,..., CPRm) for each surrogate model (SM1,..., SMm) using the compensation parameter set (CP), d) determining (S4) a weighted average friction compensation value (CPRav) of the friction compensation result values using the respective weighting factor (w1,...,wm) of the respective surrogate model (SM1,..., SMm), e) deducing (S5) a quality indicator (Q) for the friction compensation parameter set (CP) based on the weighted average friction compensation value (CPRav), f) outputting (S6) the friction compensation parameter set (CPopt), if the quality indicator (Q) fulfills a given quality criterion (QC), or repeating (S7) steps b) to e) until the quality indicator fulfills the given quality criterion, g) applying (S8) the outputted friction compensation parameter set (CPopt) to the machine tool for reducing friction within the machine tool.

Inventors:
HEIN DANIEL (DE)
UDLUFT STEFFEN (DE)
YUTKOWITZ STEPHEN (US)
Application Number:
PCT/EP2021/057457
Publication Date:
October 07, 2021
Filing Date:
March 23, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
SIEMENS AG (DE)
International Classes:
G05B19/404
Foreign References:
US20190317472A12019-10-17
US20160239594A12016-08-18
Other References:
FERREIRA WALLACE G ET AL: "Ensemble of metamodels: the augmented least squares approach", STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION, SPRINGER BERLIN HEIDELBERG, BERLIN/HEIDELBERG, vol. 53, no. 5, 15 December 2015 (2015-12-15), pages 1019 - 1046, XP035857461, ISSN: 1615-147X, [retrieved on 20151215], DOI: 10.1007/S00158-015-1366-1
Download PDF:
Claims:
Patent claims

1. Computer-implemented method for reducing friction within a machine tool (MT), comprising the method steps: a) reading (SI) a plurality of surrogate models (SM) for ap¬ proximating friction compensation within a given machine tool, wherein each surrogate model (SMI, ..., SMm) is config ured such that it assigns a friction compensation result val¬ ue to a given friction compensation parameter set for reduc ing friction within a machine tool, and wherein a weighting factor (wl, ..., wm) is assigned to each surrogate model (SMI, ..., SMm), the weighting factor representing a goodness-of-fit of a surrogate model to a machine tool, b) reading (S2) a friction compensation parameter set (CP), c) determining (S3) a friction compensation result value (CPR1, ..., CPRm) for each surrogate model (SMI, ..., SMm) using the compensation parameter set (CP), d) determining (S4) a weighted average friction compensation value (CPRav) of the friction compensation result values us¬ ing the respective weighting factor (wl,...,wm) of the respec¬ tive surrogate model (SMI, ..., SMm), e) determining (S5) a quality indicator (Q) for the friction compensation parameter set (CP) based on the weighted average friction compensation value (CPRav), f) outputting (S6) the friction compensation parameter set (CPopt), if the quality indicator (Q) fulfills a given quali¬ ty criterion (QC), or otherwise repeating (S7) steps b) to e) until the quality indicator fulfills the given quality crite¬ rion, g) applying (S8) the outputted friction compensation parame ter set (CPopt) to the machine tool for reducing friction within the machine tool.

2. Computer-implemented method according to claim 1, further comprising the steps:

- measuring (S9) a real friction compensation result value (CPR_MT) of the machine tool based on the applied friction compensation parameter set (CPopt), - modifying (S10) each weighting factor (wl,...,wm) of the re spective surrogate model (SMI, ..., SMm) depending on a dis¬ crepancy between each friction compensation result value (CPR1, ..., CPRm) and the measured real friction compensation result value resulting from the respective surrogate model and

- repeating the steps b) to g) of claim 1.

3. Computer-implemented method according to one of the pre¬ ceding claims, wherein the plurality of surrogate models (SMI, ..., SMn) is generated (SO) based on a given plurality of data sets (DATA) by means of a regression method (RM), where¬ in each data set (DATA) comprises a friction compensation pa rameter set and a corresponding friction compensation result value of a resulting reduced friction within the respective machine tool.

4. Computer-implemented method according to claim 3, wherein the data set (DATA) is generated based on a friction measure¬ ment at a real machine tool.

5. Computer-implemented method according to claim 3, wherein the data set (DATA) is generated based on a dedicated comput¬ er-aided simulation of a machine tool.

6. Computer-implemented method according to one of the pre¬ ceding claims, wherein the friction compensation parameter set (CP) is generated by means of a fitness function, wherein the fitness function depends on the surrogate models

(SMI,...SMm) and the respective weighting factor (wl,...,wm).

7. Computer-implemented method according to one of the pre¬ ceding claims, wherein the plurality of surrogate models (SM) is selected based on machine-specific identification data of the machine tool.

8. Apparatus (100) for reducing friction within a machine tool (MT), comprising: a) an input unit (101) configured to read a plurality of sur rogate models for approximating friction compensation within a given machine tool, wherein each surrogate model is config ured such that it assigns a friction compensation result val ue to a given friction compensation parameter set for reduc ing friction within a machine tool, and wherein a weighting factor is assigned to each surrogate model, the weighting factor representing a goodness-of-fit of a surrogate model to a machine tool, b) an analysis unit (102) configured to

- read in a friction compensation parameter set,

- determine a friction compensation result value for each surrogate model using the compensation parameter set,

- determine a weighted average friction compensation value of the friction compensation result values using the respective weighting factor of the respective surrogate model, and

- determine a quality indicator for the friction compensation parameter set based on the weighted average friction compen sation value, c) an output unit (103) configured to output the friction compensation parameter set, if the quality indicator fulfills a given quality criterion, or otherwise to repeat the afore mentioned steps, and d) an application unit (104) configured to apply the output ted friction compensation parameter set to the machine tool for reducing friction within the machine tool.

9. Apparatus (100) according to claim 8, wherein the applica tion unit (104) is configured to receive a measured real friction compensation result value of the machine tool based on the applied friction compensation parameter set, and the analysis unit is configured to modify each weighting factor of the respective surrogate model depending on a discrepancy between each friction compensation result value and the meas ured real friction compensation result value and resulting from the respective surrogate model and to repeat the steps b) to e) of claim 1.

10. Apparatus (100) according to claim 8 or 9 comprising a generator (105) configured to generate the plurality of sur rogate models for friction compensation based on a given plu rality of data sets (DATA) by means of a regression method (RM), wherein each data set comprises a friction compensation parameter (CP') for setting a machine tool and a correspond- ing friction compensation result (CPR/) of a resulting re duced friction within the respective machine tool.

11. Apparatus according to one of the claims 8 to 10, wherein the apparatus is connected to a database (DB), wherein the database is configured to store data sets (DATA) and/or sur rogate models (SM).

12. Computer program product directly loadable into the in ternal memory of a digital computer, comprising software code portions for performing the steps of one of the claims 1 to 7 when said computer program product is run on a computer.

Description:
Description

Reduction of friction within a machine tool

The present invention relates to a computer-implemented meth od and an apparatus for reducing friction within a machine tool.

Computerized numerical control (CNC) machines are machine tools capable of automatically producing workpieces with high precision even for complex shapes. These machine tools allow high precision manufacturing across industries. However, the friction within the machine tool, i.e., between mechanical components, leads to deviations between a controlled and an actually executed position of the tool. Such deviations due to friction can affect a targeted tolerance of a produced part.

It is therefore required to reduce the friction between me chanical parts of a machine tool. Friction compensation con trollers with tuned parameters correct errors during manufac turing introducing opposing forces. Due to variations in the forces and resulting friction of different machine parts, a controller must be set individually for each machine. Fur thermore, over the lifetime of the machine, these parameters must be recalibrated.

Conventionally, an expert technician manually tunes these pa rameters, requiring interruption of production schedules and causing machine downtime. In addition, an irregular tuning and inconsistent quality among expert technicians can lead to loss of tolerance and thereby reduced quality of finished parts.

It is therefore an objective of the present invention to im prove the friction compensation within a machine tool. The object is solved by the features of the independent claims. The dependent claims contain further developments of the invention.

The invention provides according to the first aspect a com puter-implemented method for reducing friction within a ma chine tool, comprising the method steps: a) reading a plurality of surrogate models for approximating friction compensation within a given machine tool, wherein each surrogate model is configured such that it assigns a friction compensation result value to a given friction com pensation parameter set for reducing friction within a ma chine tool, and wherein a weighting factor is assigned to each surrogate model, the weighting factor representing a goodness-of-fit of a surrogate model to a machine tool, b) reading a friction compensation parameter set, c) determining a friction compensation result value for each surrogate model using the compensation parameter set, d) determining a weighted average friction compensation value of the friction compensation result values using the respec tive weighting factor of the respective surrogate model, e) determining a quality indicator for the friction compensa tion parameter set based on the weighted average friction compensation value, f) outputting the friction compensation parameter set, if the quality indicator fulfills a given quality criterion, or oth erwise repeating steps b) to e) until the quality indicator fulfills the given quality criterion, g) applying the outputted friction compensation parameter set to the machine tool for reducing friction within the machine tool.

If not indicated differently the terms "calculate", "per form", "computer-implemented", "compute", "determine", "gen erate", "configure", "reconstruct", and the like, preferably are related to acts and/or processes and/or steps which change and/or generate data, wherein data can particularly be presented as physical data, and which can be performed by a computer or processor. The term "computer" can be interpreted broadly and can be a personal computer, server, mobile compu ting device, or a processor such as a central processing unit (CPU) or microprocessor.

The machine tool can for example be a computerized numerical control (CNC) machine tool. A surrogate model is preferably a computerized model, which is configured to approximate or fit the friction compensation behavior within a machine tool. The friction compensation, i.e. the application of for example an opposing or balancing force to reduce a frictional force, within a machine tool depends on the applied friction compen sation parameter set for controlling the machine tool.

A surrogate model can for example be a fitting model, a re gression model or an artificial neural network. A surrogate model is preferably suited to represent the physical behav ior, i.e. particularly the friction between machine parts of the machine tool due to internal forces. A friction compensa tion parameter set can be understood as input values for a surrogate model. Furthermore, a friction compensation parame ter set is input for setting a machine tool in order to re duce internal frictional forces. A weighting factor prefera bly represents the likelihood of the respective surrogate model to correctly reproduce the friction compensation re sponse within a machine tool. In other words, the weighting factor represents the goodness-of-fit of a respective surro gate model to the friction compensation response of a machine tool.

The proposed method has the advantage that an optimized fric tion compensation parameter set for setting a machine tool can be found in an automated way such that internal friction is reduced within the machine tool. Furthermore, the method can be applied on-site, i.e. in parallel with operations of the machine tool, due to computing speed. The friction compensation parameter set is used to set the machine tool such that for example an opposing force is ap plied in a way that a frictional force between machine parts is reduced. Less calibration effort to calibrate the machine tool is necessary and/or better calibration results can be achieved. Moreover, the invention can particularly be applied to unknown machines.

The invention enables determination of an optimized parameter set for friction compensation within a machine tool using a plurality of surrogate models approximating the frictional forces of preferably similar machine tools.

In a preferred embodiment of the computer-implemented method according to the first aspect of the present invention, a re al friction compensation result value of the machine tool can be measured based on the applied friction compensation param eter set, each weighting factor of the respective surrogate model can be modified depending on a discrepancy between each friction compensation result value and the measured real friction compensation result value resulting from the respec tive surrogate model and the aforementioned steps b) to g) can be repeated.

By adjusting the weighting factors of the used surrogate mod els based on the resulting friction compensation result value measured at the real machine tool, the approximation of the friction compensation for this machine tool can be optimized. Preferably, the adjustment of the weighting factors of the surrogate models and application of the friction compensation parameter is iteratively performed until a given stopping criterion is reached. Such stopping criterion can for example be a certain friction compensation quality, an expired tuning time, or the selected parameter set has been found to opti mally fit to the real machine tool.

In a further embodiment of the computer-implemented method according to the first aspect of the present invention, the plurality of surrogate models can be generated based on a given plurality of data sets by means of a regression method, wherein each data set comprises a friction compensation pa rameter set and a corresponding friction compensation result value of a resulting reduced friction within the respective machine tool.

Preferably, such data sets for generation of surrogate models for machine tools are stored in a database. Based on the available data, surrogate models can be trained by means of a regression method individually for each data set. The learn ing objective of a surrogate model is to estimate the fric tion compensation result for a given parameter set. Hence, the provided data sets can be used as training data for training these surrogate models. Possible regression tech niques are for example linear and polynomial models, regres sion trees, artificial neural networks, or Gaussian process es.

In one embodiment of the computer-implemented method accord ing to the first aspect of the present invention, a data set can be generated based on a friction measurement at a real machine tool.

In one embodiment of the computer-implemented method accord ing to the first aspect of the present invention, a data set can be generated based on a dedicated computer-aided simula tion of a machine tool.

The data sets can be stored in a database. The data sets are provided for surrogate model generation for friction compen sation within a machine tool.

In one embodiment of the computer-implemented method accord ing to the first aspect of the present invention, the fric tion compensation parameter set can be generated by means of a fitness function, wherein the fitness function depends on the surrogate models and the respective weighting factor. The friction compensation parameter set can be determined based on a computerized search of the parameter space using a fitness function. The fitness function preferably uses the surrogate models and the respective weighting factors compu ting a scalar value which serves as indicator on the quality of the used parameter set.

In one embodiment of the computer-implemented method accord ing to the first aspect of the present invention, the plural ity of surrogate models can be selected based on machine- specific identification data of the machine tool.

Preferably, the machine-specific identification data of the machine tool comprise manufacturing information data and/or machine type data. By using prior knowledge of the machine tool, the optimization process can be further improved. Pref erably, before starting an on-site optimization process for a selected machine tool, the plurality of surrogate models used to approximate the friction compensation response of the ma chine tool is preselected. Hence, additional surrogate models which are for example not well suited to approximate this ma chine tool can be excluded.

According to a second aspect, the invention relates to an ap paratus for reducing friction within a machine tool, compris ing: a) an input unit configured to read a plurality of surrogate models for approximating friction compensation within a given machine tool, wherein each surrogate model is configured such that it assigns a friction compensation result value to a given friction compensation parameter set for reducing fric tion within a machine tool, and wherein a weighting factor is assigned to each surrogate model, the weighting factor repre senting a goodness-of-fit of a surrogate model to a machine tool, b) an analysis unit configured to

- to read in a friction compensation parameter set, - determine a friction compensation result value for each surrogate model using the compensation parameter set,

- determine a weighted average friction compensation value of the friction compensation result values using the respective weighting factor of the respective surrogate model, and

- determine a quality indicator for the friction compensation parameter set based on the weighted average friction compen sation value, c) an output unit configured to output friction compensation parameter set, if the quality indicator fulfills a given quality criterion, or otherwise to repeat the steps performed by the analysis unit, and d) an application unit configured to apply the outputted friction compensation parameter set to the machine tool for reducing friction within the machine tool.

The apparatus is preferably connected to the machine tool or it is part of the machine tool. The apparatus and/or at least one of its units can further comprise at least one processor or computer to perform the method steps according to the in vention. A respective unit may be implemented in hardware and/or in software. If said unit is implemented in hardware, it may be embodied as a device, e.g. as a computer or as a processor or as a part of a system. If said unit is imple mented in software it may be embodied as a computer program product, as a function, as a routine, as a program code or as an executable object. The output unit preferably provides a data structure comprising the optimized compensation parame ter. Such data structure can be for example transmitted to a control unit of the machine tool for setting the machine tool accordingly .

According to an embodiment of the apparatus, the application unit can be further configured to receive a measured real friction compensation result value of the machine tool based on the applied friction compensation parameter set, and the analysis unit is configured to modify each weighting factor of the respective surrogate model depending on a discrepancy between each friction compensation result value and the meas ured real friction compensation result value and resulting from the respective surrogate model and to repeat the steps b) to e) of the computer-implemented method according to the first aspect of the invention.

According to an embodiment, the apparatus can comprise a gen erator which is configured to generate the plurality of sur rogate models for friction compensation based on a given plu rality of data sets by means of a regression method, wherein each data set comprises a friction compensation parameter for setting a machine tool and a corresponding friction compensa tion result of a resulting reduced friction within the re spective machine tool.

According to a further embodiment, the apparatus can be con nected to a database, wherein the database is configured to store data sets and/or surrogate models.

The invention further comprises a computer program product directly loadable into the internal memory of a digital com puter, comprising software code portions for performing the steps of the said method when said product is run on a com puter.

A computer program product, such as a computer program means, may be embodied as a memory card, USB stick, CD-ROM, DVD or as a file which may be downloaded from a server in a network.

The invention will be explained in more detail by reference to the accompanying figures.

Fig. 1 shows a flow chart including method steps involved in an embodiment of the method for reducing fric tion within a machine tool; Fig. 2 shows a schematic representation of an embodiment of the method for reducing friction within a ma chine tool; and

Fig. 3 shows a schematic representation of an embodiment of an apparatus for reducing friction within a ma chine tool.

Equivalent parts in the different figures are labeled with the same reference signs.

Figure 1 shows a flow chart illustrating the method steps in volved in a computer-implemented method for reducing friction within a machine tool, preferably a CNC machine tool. The method provides preferably an optimized parameter set which can be applied to the machine tool to achieve optimal fric tion compensation. Furthermore, the method allows to deter mine an optimized surrogate model to approximate friction compensation within the machine tool. The machine tool can be for example a CNC machine tool for milling, laser cutting, punch pressing or other industrial application.

The first step SO of the method involves generation of a plu rality of surrogate models for different machine tools by means of a regression method based on training data. The training data comprises data sets. Each data set includes a friction compensation parameter set, further also called a parameter set, for setting a machine tool, and a correspond ing friction compensation result value. A friction compensa tion result value results from applying the parameter set to the machine tool and measuring the resulting frictional forc es. Hence, the friction compensation result value can be un derstood as an indicator of the resulting frictional forces within the machine tool, i.e. it can for example be deter mined based on sensor measurements measuring the friction be tween machine parts. The data sets can be generated based on friction measurements at a real machine tool or different machine tools and/or based on a dedicated computer-aided simulation of at least one machine tool. A respective surrogate model for a machine tool is generated based on at least one data set. For genera tion of the surrogate models a computerized regression method can be used, as for example linear or polynomial models, re gression trees, artificial neural networks or Gaussian pro cesses. The surrogate models are preferably generated for a plurality of different machine tools. The generated surrogate models are preferably stored in a database.

In the next step SI a plurality of surrogate models for ap proximating friction compensation within machine tools is read in. Preferably, from the available surrogate models stored in the database, a sample of surrogate models is se lected based on machine-specific identification data of the machine tool, e.g. machine type. A weighting factor is as signed to each surrogate model, wherein the weighting factor preferably represents a goodness-of-fit of the respective surrogate model approximating the friction compensation re sponse of the real machine tool. At start of the optimization procedure, the weighting factor of each surrogate model can particularly be equally distributed, as e.g. all set to 1.

In the next step S2 a friction compensation parameter set is read in. Preferably, the friction compensation parameter set is generated based on given weighting criteria, which is ex plained below. In general, the friction compensation parame ter set preferably comprises at least one parameter to con trol the machine tool, which is also input for a respective surrogate model. The friction compensation parameter set can be proposed based on evaluation of the surrogate models, as explained below. The initially proposed parameter set can for example be an initial estimate.

In the next step S3, based on the inputted friction compensa tion parameter set, friction compensation result values are determined for each inputted surrogate model. In other words, each surrogate model is evaluated to provide a friction com pensation result value based on the parameter set.

In the next step S4, a weighted average is determined based on the compensation result values and the weighting factors of the corresponding surrogate models.

In the next step S5, a quality indicator of the used friction compensation parameter set is determined based on the weighted average compensation result value. The quality indi cator represents the quality of the proposed parameter set for reducing friction when applied to the machine tool. The quality indicator can for example have the value of the cor responding weighted average compensation result value and/or multiplied by a given factor or similar.

If the quality indicator fulfills a given quality criterion, e.g. exceeding a given threshold value, the friction compen sation parameter set is outputted, step S6, and applied to the machine tool, step S8, for setting the machine tool in such a way that internal friction is reduced. The friction compensation parameter set can for example be transmitted to a machine control unit for controlling the machine tool in order to reduce friction between mechanical parts of the ma chine tool.

If the quality indicator does not fulfill the given quality criterion, step S7, a second friction compensation parameter set, which differs from the first inputted parameter set, is selected and inputted. Using this second parameter set, sec ond friction compensation result values are determined for the surrogate models. The weighting factors of the surrogate models are preferably not modified. A weighted average value of the resulting second friction compensation result values is determined to deduce a quality indicator for the second parameter set. If the quality indicator of the second parame ter set meets the given quality criterion, the second fric- tion compensation parameter set is outputted. If it does not meet the quality criterion, the search for a suitable parame ter set is repeated. Hence, a suitable parameter set is searched based on this iterative process. In particular, this parameter search can be implemented using a fitness function which uses the surrogate models and the respective weighting factors of the surrogate models.

The friction compensation parameter set which meets the qual ity criterion is applied to the machine tool, step S8, and a real friction compensation result value can be measured based on the applied parameter set, step S9. In the next step S10, the weighting factors of the surrogate models can be modified depending on the discrepancy between the measured real fric tion compensation result and each friction compensation re sult value outputted by each surrogate model. For example, a small difference between the measured and one modelled fric tion compensation result can transform into a higher weighting of the respective surrogate model. Based on the modified weighting factors of the surrogate models, steps S2 to S8, and preferably to S10, can be repeated, further im proving modelling and parameter determination for the machine tool.

Figure 2 shows schematic representation of an embodiment of the method for reducing friction within a machine tool MT.

The representation comprises the surrogate model generation by means of a generator 103. A model generator 103 comprises preferably a database DB or is connected to a database DB.

The database DB preferably comprises training data DATA, which are used to generate surrogate models which are suited to approximate and reproduce frictional forces within machine tools. The training data DATA are based on measurement data and/or simulation data. The training data DATA comprise a plurality of data sets wherein each data set consists of a friction compensation parameter set CP' and a corresponding friction compensation result value CPR/ . Using the training data DATA, the generator 103 can generate a plurality of sur- rogate models SMI, ..., SMn by means of a regression method RM. The surrogate models SMI, ..., SMn can be stored in the data base DB.

At least a sample of surrogate models SMI, ..., SMm is selected from this plurality of surrogate models SMI, ..., SMn. The se lection is preferably based on machine-specific identifica tion data of the machine tool MT. A weighting factor wl, ..., wm is assigned to each surrogate model SMI, ..., SMm. A weighting factor wl, ..., wm represents the goodness-of-fit of the respective surrogate model approximating the friction re sponse of the machine tool.

The selected surrogate models SMI, ..., SMm are read in by the analysis unit 102. Furthermore, one friction compensation pa rameter set CP is read in by the analysis unit 102. Prefera bly, the friction compensation parameter set CP is determined using a fitness function based on the selected surrogate mod els SMI, ..., SMm and their respective weighting factors wl, ..., wm.

For each surrogate model SMI, ..., SMm a corresponding friction compensation result value CPR1, ..., CPRm, is determined based on the inputted parameter set CP. Using the respective weighting factors wl,...,wm, a weighted average value CPRav of these friction compensation result values CPR1, ..., CPRm is calculated. From this weighted average CPRav a quality indi cator Q is deduced to determine the matching quality of the friction compensation parameter set CP. If the quality indi cator meets a given quality criterion QC, the friction com pensation parameter CP is outputted as an optimized friction compensation parameter CPopt and applied to the machine tool MT. Otherwise, another parameter set can be proposed and evaluated until a parameter set meets the quality criterion QC.

At the machine tool MT, a real friction compensation result value CPR_MT can be measured. For example, frictional forces between two machine parts can be measured using a sensor. By comparing this measured friction compensation result value CPR_MT with the individual friction compensation result val ues CPR1, ..., CPRm outputted by the surrogate models SMI, ..., SMm, the weighting factors wl, ..., wm of these surrogate mod els can be adjusted. In other words, the weighting factors of the respective surrogate models are modified based on the fitting quality of the respective model. Preferably, more weight is given to the surrogate models SMI, ..., SMm which predict a compensation result value close to the measured one. Performing these iterative steps further improves the surrogate model weighting as well as the parameter set search, resulting in reduced friction within the machine tool. The iterative parameter search and/or model weighting can be stopped as soon as a given stopping criterion is reached.

Figure 3 shows a schematic representation of an embodiment of an apparatus 100 for reducing friction within a machine tool MT. The apparatus 100 is preferably connected to the machine tool MT using a wireless or wired connection.

The apparatus comprises an input unit 101 configured to read a plurality of surrogate models for approximating friction compensation within a given machine tool. Each surrogate mod el is configured such that it assigns a friction compensation result value to a given friction compensation parameter set for reducing friction within a machine tool. A weighting fac tor is assigned to each surrogate model.

The apparatus 100 further comprises an analysis unit 102 con figured to read in a friction compensation parameter set and to determine a friction compensation result value for each surrogate model using the compensation parameter set. The analysis unit 102 is further configured to determine a weighted average friction compensation value of the friction compensation result values using the respective weighting factor of the respective surrogate model and to deduce a quality indicator for the friction compensation parameter set based on the weighted average friction compensation value.

The apparatus 100 further comprises an output unit 103 con figured to output the friction compensation parameter set, if the quality indicator fulfills a given quality criterion, and an application unit 104 configured to apply the outputted friction compensation parameter set to the machine tool for reducing friction within the machine tool.

The application unit 104 can further be configured to receive a measured friction compensation result value of the machine tool based on the applied friction compensation parameter set. The measurement can for example be performed by means of a sensor at or inside the machine tool. The analysis unit 102 can be configured to modify each weighting factor of the re spective surrogate model depending on a discrepancy between each friction compensation result value and the measured real friction compensation result value and resulting from the re spective surrogate model and to repeat parameter set selec tion steps.

The apparatus 100 can further comprise a generator 105 con figured to generate the plurality of surrogate models for friction compensation based on a given plurality of data sets by means of a regression method, wherein each data set com prises a friction compensation parameter for setting a ma chine tool and a corresponding friction compensation result of a resulting reduced friction within the respective machine tool. Alternatively, the generator 105 can be installed sepa rately and connected to the apparatus 100.

The apparatus 100 and/or the generator 105 can further be connected to a database DB, wherein the database is config ured to store surrogate models and/or friction compensation data for generation of surrogate models for approximating friction compensation within a machine tool. Although the present invention has been described in detail with reference to the preferred embodiment, it is to be un derstood that the present invention is not limited by the disclosed examples, and that numerous additional modifica- tions and variations could be made thereto by a person skilled in the art without departing from the scope of the invention.