# US 2021/0268620 A1 · Automated drilling optimization method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0268620 A1
- Document ID: US20210268620A1
- Status: Published application
- Kind: A1
- Application number: 17/187,212
- Filed: 2021-02-26
- Published: 2021-09-02
- Earliest claimed: 2020-02-27
- Assignees: Airbus SAS; Airbus Operations S.L.
- Applicants: Airbus SAS; Airbus Operations S.L.
- Inventors: Christophe LOUAT; Arnaud ZANNI; Juan Carlos SOUTHWOOD CARRASCO
- Web page: https://patentgrep.com/patent/US20210268620A1

## CPC

- B23Q 15/22: Control or regulation of position of tool or workpiece
- B23Q 15/24: of linear position
- B23Q 15/26: of angular position
- B23Q 17/2233: {for adjusting the tool relative to the workpiece}
- G05B 13/04: involving the use of models or simulators
- G06N 20/00: Machine learning
- B23Q: DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING (tools of the kind used in lathes or boring machines B23B27/00); MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method for optimizing the execution of automated drilling systems controlled by a numerical control, NC, machine, wherein the NC machine performs the following steps: (i) identifying each drill executed by an automated drilling tool at first drilling event (t); storing a theoretical position of each executed drill at the event (t); (iii) calculating a learned position using a machine learning model and based on the stored theoretical position at the event (t); (iv) estimating an intermediate position by applying a tendency statistical function, and (v), if the difference between the intermediate position and the learned position is less than a pre-configured threshold, using the intermediate position for the drilling tool to position a next drill and execute it at a subsequent event (t+1).

## Claims

### Claim 1

A method execution of an automated machining process using a numerical control machine comprising:
i) identifying, with a unique identifier defined by the numerical control machine, to a drill process by a drilling tool at an event (t);
ii) storing a theoretical position of the drill process executed at the event (t), wherein the theoretical position is a pre-configured static position;
iii) calculating a learned position using a machine learning model and based on the stored theoretical position for the event (t);
iv) estimating an intermediate position by applying a statistical tendency function, and
if a difference between the intermediate position and the learned position is less than a pre-configured threshold, using the intermediate position to position the drilling tool to execute another drill process at subsequent event (t+1); and
if the difference between the intermediate position and the learned position is greater than the pre-configured threshold, storing the intermediate position for the subsequent event (t+1) and repeating the steps iii) and iv).

### Claim 2 (depends on 1)

The method according to claim 1, wherein the steps i) to iv) are performed automatically by the numerical control machine.

### Claim 3 (depends on 1)

The method according to claim 1, wherein the statistical tendency function is a regression algorithm.

### Claim 4 (depends on 1)

The method according to claim 1, wherein the step iii) of calculating the learned position at the event (t) comprises:
calculating a normal vector for the learned position at the event (t), wherein the normal vector is orthogonal to a surface on a workpiece to be drilled, and the normal vector is defined by components It, Jt and Kt;
calculating, using a machine learning model, a correction vector for the event (t), wherein the machine learning model is defined by components Correction_It, Correction_Jt and Correction_Kt, wherein the machine learning model combines a first machine learning model for the component It, a second machine learning model for the component Jt and a third machine learning model for the component Kt, and
wherein the machine learning model receives as input data comprising a theoretical vector defined by Theoretical_Xt, Theoretical_Yt and Theoretical_Zt, which are coordinates in the three dimensional space defining the theoretical position of the drill to be executed at the event (t);
the first machine learning model returns as output the first component Correction_It of the correction vector;
the second machine learning model returns as output the second component Correction_Jt of the correction vector; and
the third machine learning model returns as output the third component Correction_Kt of the correction vector,
wherein the learned position being defined in the three dimensional space by the coordinates Xt, Yt and Zt.

### Claim 5 (depends on 4)

The method according to claim 4, wherein the input data taken by the machine learning further comprises a correction vector calculated at a prior event (t−1) with components Correction_It-1, Correction_Jt-1 and Correction_Kt-1.

### Claim 6 (depends on 4)

The method according to claim 4, wherein the components of the correction vector are calculated in degrees of angular movement for the automated drilling tool.

### Claim 7 (depends on 4)

The method according to claim 4, further comprising computing by the numerical control machine an error for the event (t) corresponding to a difference between the intermediate position and the learned position and determining whether the error is under the pre-configured threshold, the error being computed as
|Correction_It−Correction_It-1| for the first machine learning model,
|Correction_Jt−Correction_Jt-1| for the second machine learning model and
|Correction_Kt−Correction_Kt-1| for the third machine learning model.
