Tuning of axis control of multi-axis machines

FIELD OF THE INVENTION

The present invention relates to tuning of axis control of multi-axis machines for particular use in materials pro cessing. Various embodiments of the invention relate to sys tems for tuning of axis control of multi-axis machines and to corresponding methods of operating such systems.

BACKGROUND OF THE INVENTION

Machining relates to a manufacturing process where computer numerically controlled, CNC, machine tools are used to shape pieces of rigid material, i.e., workpieces, into desired shapes and sizes by subtractive manufacturing, i.e. by con trolled material removal, for example by cutting or milling. Machining has high requirements with respect to, inter alia, surface quality, accuracy and cutting time of a workpiece.

Multi-axis machining relates to machining based on multiple axes which are associated with respective spatial degrees of freedom. Each axis of movement is implemented either by mov ing a table to which a workpiece is attached, or by moving a tool. The multiple axes facilitate achievement of the above- referenced requirements and enable manufacturing of more com plex workpieces as well. in order to enable the axes to move according to the machine requirements, the axes' feedback loops as well as additional features of computer numerical control and drive control, such as jerk limitation or friction compensation, have to be tuned in a corresponding way. There are many different con trol features available which are either used to compensate a mechanical property of an axis' drive train, or for achieving specific tuning goals, such as high speed cutting or mold & die. Each feature has various modes and parameters to face different mechanical situations and different requirements.

In addition, there are features with conflicting goals. To achieve the desired quality of motion, i.e. accuracy, point- to-point time, and without excitation or vibration, the rele vant features have to be selected, the relevant mode and pa rameters have to be determined with respect to properties of real axis and according to the machine requirements.

Multi-axis machining requires an experienced engineer to se lect the relevant features and test the effectiveness at the real axis. Furthermore, the most expedient mode of this fea ture has to be found, either by reading documentation, or by trial-and-error . The adaption of the feature parameters to the specific axis properties and tuning of theses parameters with respect to the machine requirement is a time consuming task with many iterations. If the tuning has not been suc cessful, it is a difficult task to find out, whether

- the combination of features is correct,

- all relevant features have been used,

- the corresponding feature modes have been used,

- there is a chance to affect the workpiece quality issue by axis tuning,

- there is a chance to solve the axis quality of movement issue by applying control features.

Today, only a few domain experts / engineers at the machine manufacturer are responsible for and capable of tuning the corresponding control and drive settings. They have to gather information about the features and obtain experience in ana lyzing problems. Furthermore, the responsible expert has to decide whether the workpiece quality and machine requirements can successfully be achieved under the current circumstances, such as machine axis properties, available control features and available feature modes, by adapting control parameters. In addition to the time consumption and effort for tests and measurements, it is uncertain how much improvement in terms of cutting time or workpiece quality is possible. If the en- gineer fails to memorize or report the findings of his analy sis, no data / experience from these tuning results can be reused. Furthermore, no classification of axes types, results and proved methods is possible.

Known solutions are based on raw data collected by individual users .

BRIEF SUMMARY OF THE INVENTION

In view of the above, there is a need in the art for systems for tuning of axis control of multi-axis machines and for methods of operating such systems, which address some of the above needs .

These underlying objects of the invention are each solved by the system and the method as defined by the independent claims. Preferred embodiments of the invention are set forth in the dependent claims.

According to a first aspect, a system for tuning of axis con trol of a multi-axis machine is provided. The system compris es a knowledge base for acquiring and maintaining factual knowledge associated with the tuning of the axis control. The factual knowledge has a uniform ontology and a uniform data representation. In addition, the factual knowledge comprises known input facts and associated known output facts. The sys tem further comprises an inference unit for automatically in ferring new output facts associated with given new input facts in accordance with the factual knowledge.

Automatically inferring new output facts associated with giv en new input facts in accordance with the factual knowledge facilitates

- obtaining a tuning recommendation with higher rate of suc cess,

- reducing time for tuning by reducing number of iterations, - obtaining a proposal for expected improvement when using a known tuning strategy, and

- obtaining expected values for each step (strategy, modes, measurement conditions, results) .

In addition, the uniform ontology and the uniform data repre sentation facilitate acquisition of factual knowledge from a number of sources, which may have used different methodolo gies and / or metrics in their data collection.

The term "tuning" as used herein may refer to optimization of systems and methods, especially under computer numerical con trol, according to given tuning requirements / goals.

The term "computer numerical control", or "CNC", may refer to automated control of machining tools by means of a computer, to process a workpiece according to given requirements.

The term "axis control" as used herein may refer to a control of feedback loops and / or numerical control features of ax es, including a speed control of the actual drives / motors, for example, and excluding a basic voltage and/or current control thereof.

The term "ontology" as used herein may refer to a system of information having logic relations. For example, an ontology may define which information entities exist and how such en tities may be grouped, related within a hierarchy, and subdi vided according to similarities and differences.

The term "uniform data representation" as used herein may re fer to a circumstance that similar data is represented in a same way, both syntactically and semantically. As a syntactic example, real-valued numbers may be represented in a uniform floating-point data format. As a further example, character strings may be represented in a uniform character string data format, and so on. It may be useful to provide identifiers which impart the particular data format or representation in use. In particular, if the data is collected from a number of sources, which may have used different methodologies and / or metrics in their data collection, it may be required to per form data conversion into the appropriate uniform data repre sentation. As semantic examples, uniform figures of merit may be defined for rating of requirements, of a dynamic behavior of the multiple axes and / or of a tuning result. For in stance, it may be desirable to describe

- a degree of tuning success, which may be a trade-off be tween a measured surface quality, measured contour accura cy, and measured processing time,

- a degree of accuracy of a measurement,

- a degree of challenge of a tuning result, and

- a degree of difficulty of overall or individual require ment fulfillment.

The axis control of each of the multiple axes may respective ly be associated with a number of available axis control fea tures, a number of available axis control modes associated with the number of available axis control features, and a number of available axis control parameters associated with the number of available axis control modes.

The input facts may respectively comprise at least one of a tolerable surface quality, a tolerable accuracy, and a toler able machining time.

The input facts may further comprise uniform, machine- readable descriptions of

- machine types, such as three-axes vertical machining cen ter with one main spindle, five-axis machining center with real five-axis cutting, lathe, mill-turn machining center, etc . ,

- machine axis types, such as single axis, coupled axis,

horizontal / vertical / rotary mechanical drive train with gear box, leadscrew or directly driven, etc.,

- application types, such as milling, grinding, turning, la ser cutting, - machine requirements, and

- workpiece requirements, such as a machined material - in particular a metal.

The term "tolerable surface quality" as used herein may refer to a tolerable degree of non-conformity of a shaped surface to a perfectly shaped surface, such as a perfectly flat sur face, for example.

The term "tolerable accuracy" as used herein may refer to a tolerable degree of non-conformity of a measure of dimension to a predefined specification of this measure.

The term "tolerable machining time" as used herein may refer to a tolerable time period required to at least partially shape a workpiece, or a detail of the same, using a same set of output facts.

The output facts may respectively comprise at least one of the number (i.e. one or more) of available axis control fea tures, at least one of the number (i.e. one or more) of available axis control modes associated with each of the at least one of the number of available axis control features, at least one of the number (i.e. one or more) of available axis control parameters associated with each of the at least one of the number of available axis control modes, and ac ceptable values for each the at least one of the number of available axis control parameters.

The output facts may further comprise uniform, machine- readable descriptions of

- machine types, such as three-axes vertical machining cen ter with one main spindle, five-axis machining center with real five-axis cutting, lathe, mill-turn machining center, etc . ,

- machine axis types, such as single axis, coupled axis,

horizontal / vertical / rotary mechanical drive train with gear box, leadscrew or directly driven, etc., - application types, such as milling, grinding, turning, la ser cutting,

- machine requirements,

- workpiece requirements,

- reports and automatically saved data about successful tun ing experiments,

- saved real properties of machine axes, and methods to de rive the same,

- criteria of an importance of an axis for the machine pro cess and its requirements,

- criteria of an impact of an axis for a required property of the machine process or related requirements,

- criteria of successful tuning:

a. measured processing time, or improvement of the same, b. measured accuracy, or improvement of the same, and c. measured surface quality, or improvement of the same.

Saving a combination of the relevant ones of the above infor mation objects together with an evaluation of success reduces trial-and-error when tuning axis control of multi-axis ma chines .

The term "measured surface quality" as used herein may refer to a measured degree of non-conformity of a shaped surface to a perfectly shaped surface, such as a perfectly flat surface, for example.

The term "measured accuracy" as used herein may refer to a measured degree of conformity of a measure of dimension to a predefined specification of this measure.

The term "measured machining time" as used herein may refer to a measured time period required to at least partially shape a workpiece, or a detail of the same, using a same set of output facts.

Contradicting output facts may be prevented. For example, such contradiction may be prevented by mutual exclusion or limitation of interacting output facts according to known heuristics, for instance encoded as if-then-else case distinctions.

This reduces trial-and-error when tuning the axes' control of multi-axis machines.

The knowledge base may be configured for acquiring the factu al knowledge from a plurality of manufacturing or test cases on a plurality of multi-axis machines.

This facilitates solving the so-called "knowledge acquisition problem", which denotes the problem of acquiring available factual knowledge from usually rare and expensive domain ex perts. As such experts are potentially affiliated with dif ferent divisions of an organization and / or different organ izations, a broader base of factual knowledge may be acquired to draw from, e.g. for re-use and further inspection.

The inference unit may be configured for automatically infer ring, in accordance with the factual knowledge, the new out put facts associated with the new input facts, if the new in put facts represent interpolations or extrapolations of some of the known input facts.

Particularly, there may be cases wherein the new input facts deviate from the some of the known input facts only in de tails. This may, for example, relate to a deviation in as few as a single information item. In such cases, it may be possi ble to map the some of the known input facts to the new input facts by interpolation or extrapolation based on an inherent relation, such as determinable by regression analysis or heu ristics, for example. Accordingly, a tuning parameter may be explored by only a carefully selected small number of test cases . If, however, the new input facts are identical to any of the known input facts, it may be possible to re-use the latter for tuning directly.

The term "regression analysis" as used herein may relate to a set of statistical processes for estimating the relationships among variables, and in particular between a dependent varia ble and one or more independent variables.

The term "heuristics" as used herein may relate to any ap proach of problem solving that employs a practical method to yield a satisfactory solution, especially if finding an opti mal solution is impossible or impractical.

The inference unit may be configured for automatically infer ring, in accordance with the factual knowledge, the new out put facts associated with the new input facts based on the associations between the known input facts and the known out put facts.

Automatically inferring, in accordance with the factual knowledge, the new output facts associated with the new input facts facilitates

- giving proposals for a tuning strategy:

a. which features and tuning steps are recommended, b. which modes of these features are recommended, and / or

c. a measurement strategy for each recommended step.

- giving proposals for a tuning result, if the properties of the real axis and the found cloud data match very well,

- giving proposals for an expected range of tuning result, such as an expected value and measure of variance,

- giving an initial judgement of the current axis tuning

status - before further tuning - with respect to the known facts ,

- giving an automatic judgement of a measurement result with respect to the known facts, and - giving an automatic judgement of a single tuning experi ment result with respect to the known facts.

The system may further comprise a learning unit for automati cally learning the associations between the known input facts and the known output facts.

This facilitates capturing the inherent associations between the input facts and the output facts directly, for instance without any detours via interpolation or extrapolation, and without a need to capture these inherent associations explic itly.

The learning unit may be configured for automatically learn ing the associations between the known input facts and the known output facts based on machine learning.

This facilitates capturing the inherent associations between the input facts and the output facts directly, without a need to engineer such a capturing procedure explicitly.

For example, machine learning may be used to

- classify machine axis types, machine axes properties, as well as typical combinations, and

- improve the learned associations between the known input facts and the known output facts based on user feedback, potentially based on automatic test cases carried out by real engineers, if desired.

The term "machine learning" as used herein may relate to giv ing computer systems the ability to progressively improve performance on a specific task based on given data, such as the above-referenced known factual knowledge, without being explicitly programmed. Machine learning algorithms may learn from and make predictions on the given data, and may overcome following strictly static program instructions by making da ta-driven predictions or decisions through building a model from sample inputs. For example, machine learning may be based on deep learning using artificial neural networks hav ing multiple hidden layers.

The learning unit may be configured for automatically learn ing the associations between the known input facts and the known output facts based on supervised machine learning using the factual knowledge.

The term "supervised machine learning" as used herein may re late to presenting computer systems with example inputs and their desired outputs in order to learn a general rule that maps the example inputs to the desired outputs. In the case of tuning of machine axes, the example inputs, desired out puts and general rules correspond to known input facts, known output facts, and their associations in between.

According to a second aspect, a method of operating a system for tuning of axis control of a multi-axis machine is provid ed. The method comprises a step of acquiring and maintaining factual knowledge associated with the tuning of the axis con trol. The factual knowledge has a uniform ontology and a uni form data representation. In addition, the factual knowledge comprises known input facts and associated known output facts. The method further comprises a step of automatically inferring new output facts associated with given new input facts in accordance with the factual knowledge.

The method may be used to operate the system of various em bodiments .

Advantageously, the technical effects and advantages de scribed above in relation with the systems for tuning of axis control of a multi-axis machine equally apply to the corre sponding methods having corresponding features.

BRIEF DESCRIPTION OF THE DRAWINGS Embodiments of the invention will be described with reference to the accompanying drawings, in which the same or similar reference numerals designate the same or similar elements.

Fig. 1 is a schematic view of a system 10 according to an em bodiment for tuning of axis control of a multi-axis machine 30.

Fig. 2 is a schematic view of an ontology 21 of the factual knowledge 20 associated with the tuning of the axis control of a multi-axis machine 30.

Fig. 3 is a schematic view of a cloud-based system 10 accord ing to an embodiment in connection with a plurality of multi-axis machines 30.

Fig. 4 is a schematic view of a method 40 according to an em bodiment of operating a system 10 of an embodiment for tuning of axis control of a multi-axis machine 30.

DETAILED DESCRIPTION OF EMBODIMENTS

Exemplary embodiments of the invention will now be described with reference to the drawings. While some embodiments will be described in the context of specific fields of applica tion, the embodiments are not limited to this field of appli cation. Further, the features of the various embodiments may be combined with each other unless specifically stated other wise .

The drawings are to be regarded as being schematic represen tations and elements illustrated in the drawings are not nec essarily shown to scale. Rather, the various elements are represented such that their function and general purpose be come apparent to a person skilled in the art. Fig. 1 is a schematic view of a system 10 according to an em bodiment for tuning of axis control of a multi-axis machine 30.

With reference to Fig. 1, it will be appreciated that a sys tem 10 for tuning of axis control of a multi-axis machine 30 may comprise a knowledge base 11, an inference unit 12, as well as a learning unit 13.

The knowledge base 11 is configured for acquiring 41 and maintaining factual knowledge 20 associated with the tuning of the axis control, which factual knowledge 20 has a uniform ontology 21 and a uniform data representation, and comprises known input facts 211 associated with known output facts 212.

The inference unit 12 is configured for automatically infer ring 42 new output facts associated with given new input facts in accordance with the factual knowledge 20.

For example, the inference unit 12 may be configured for au tomatically inferring 42, in accordance with the factual knowledge 20, the new output facts associated with the new input facts, if the new input facts represent interpolations or extrapolations of some of the known input facts 211.

Alternatively or additionally, the inference unit 12 may be configured for automatically inferring 42, in accordance with the factual knowledge 20, the new output facts associated with the new input facts based on the associations between the known input facts 211 and the known output facts 212.

To this end, the learning unit 13 is configured for automati cally learning 43 the associations between the known input facts 211 and the known output facts 212, in order to gener alize the associations and to be able to apply the general ized associations to new input facts as well, resulting in new output facts. For example, the learning unit 13 may be configured for auto matically learning 43 the associations between the known in put facts 211 and the known output facts 212 based on machine learning, in particular based on supervised machine learning using the factual knowledge 20.

Fig. 2 is a schematic view of a possible ontology 21 of the factual knowledge 20 associated with the tuning of the axis control of a multi-axis machine 30.

With reference to Fig. 2, it will be appreciated that the factual knowledge 20 maintained by the knowledge base 11 has a uniform ontology 21, a uniform data representation, and comprises known input facts 211 shown at the right-hand side of Fig. 2 and associated with known output facts 212 shown at the left-hand side of Fig. 2.

The input facts - which are either known 211 or new - respec tively comprise at least one of a machining method, a machin ing tool, a machined material, a tolerable surface quality, a tolerable accuracy, and a tolerable machining time.

As will be further appreciated with reference to Fig. 2, the output facts 212 of the factual knowledge 20 have a uniform ontology 21 defining which possible information entities ex ist and how such entities may be grouped, related within a hierarchy, and subdivided according to similarities and dif ferences. The axis control of each of the multiple axes is associated with the following information entities:

- a number of available axis control features 213,

- a number of available axis control modes 214 associated with the number of available axis control features 213, and

- a number of available axis control parameters 215 associ ated with the number of available axis control modes 214.

Accordingly, the output facts - which are either known 212 or new - respectively comprise - at least one of the number of available axis control fea tures 213,

- at least one of the number of available axis control modes 214 associated with each of the at least one of the number of available axis control features 213,

- at least one of the number of available axis control pa rameters 215 associated with each of the at least one of the number of available axis control modes 214, and

- acceptable values (not shown) for each of the at least one of the number of available axis control parameters 215.

A contradiction within the respective output facts - for ex ample due to conflicting goals - is prevented.

Fig. 3 is a schematic view of a cloud-based system 10 accord ing to an embodiment in connection with a plurality of multi axis machines 30.

With reference to Fig. 3, it will be appreciated that the system 10, and more specifically its knowledge base 11, is configured for acquiring 41 the factual knowledge 20 from a plurality of multi-axis machines 30.

To this end, the system 10 is in communication with the plu rality of multi-axis machines 30 via a wire-less and/or wire- bound network infrastructure 31, which is indicated in Fig. 3 as a cloud.

A plurality of manufacturing or test cases may be carried out on the plurality of multi-axis machines 30, yielding a broad base of factual knowledge 20 to draw from.

Fig. 4 is a schematic view of a method 40 according to an em bodiment of operating a system 10 of an embodiment for tuning of axis control of a multi-axis machine 30.

With reference to Fig. 4, it will be appreciated that the method 40 may comprise the steps of acquiring 41 and main- taining, automatically inferring 42, and automatically learn ing 43.

At step 41, factual knowledge 20 associated with the tuning of the axis control is being acquired 41 and maintained, which factual knowledge 20 has a uniform ontology 21, a uni form data representation and comprises known input facts 211 associated with known output facts 212.

At step 42, new output facts associated with given new input facts are being automatically inferred 42 in accordance with the factual knowledge 20.

At step 43, the associations between the known input facts 211 and the known output facts 212 are being automatically learned 43.

The method 40 may be used to operate the system 10 of various embodiments .

While systems 10 and methods 40 according to various embodi ments have been described, various modifications may be im plemented in other embodiments. For illustration, machine learning may be based on various approaches such as genetic algorithms, reinforcement learning, and the like.

The present invention may, of course, be carried out in other ways than those specifically set forth herein without depart ing from essential characteristics of the invention. For ex ample, the previous embodiments described the present inven tion in a context of tuning of axis control of multi-axis ma chines, and for particular use in materials processing. How ever, those skilled in the art will appreciate that the pre sent invention is not so limited. The present invention may also be used in other contexts as well which require complex control tasks being tuned by expensive domain experts. The present embodiments are therefore to be considered in all re spects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the ap pended claims are intended to be embraced therein.